Traditional camera sensors rely on human eyes for observation. However, the human eye 1 is prone to fatigue when observing targets of different sizes for a long time in complex scenes, and 2 human cognition is limited, which often leads to judgment errors and greatly reduces the efficiency. 3 Target recognition technology is an important technology to judge the target category in camera 4 sensor. In order to solve this problem, a small size target detection algorithm for special scenarios was 5 proposed by this paper. Its advantage is that this algorithm not only has higher precision for small 6 size target detection, but also can ensure that the detection accuracy of each size is not lower than the 7 existing algorithm. In this paper, a new down-sampling method was proposed, which could better 8 preserve the context feature information. The feature fusion network was improved to effectively 9 combine shallow information and deep information. A new network structure was proposed to 10 effectively improve the detection accuracy of the model. In terms of accuracy, it is better than: YOLOX, 11 YOLOXR, YOLOv3, scaled YOLOv5, YOLOv7-Tiny and YOLOv8.Three authoritative public data sets 12 were used in this experiment: a) On Visdron data sets (small size targets), DC-YOLOv8 is 2.5% more 13 accurate than YOLOv8. b) On Tinyperson data sets (minimal size targets), DC-YOLOv8 is 1% more 14 accurate than YOLOv8. c) On PASCAL VOC2007 data sets (Normal size target), DC-YOLOv8 is 0.5% 15 more accurate than YOLOv8.

Pedestrian multi-target tracking technology plays an important role in artificial intelligence, driverless, virtual reality and other fields. The pedestrian multi-target tracking algorithm DeepSORT based on detection is widely used in industry. It mainly tracks multiple pedestrian targets continuously and keeps their ID unchanged. In order to improve the applicability and tracking accuracy of DeepSORT algorithm, this paper improved the IOU distance measurement in the matching process. At the same time, ResNet50 is used as the feature extraction backbone network, and combined with FPN (Feature Pyramid Network), the appearance features of multi-layer pedestrians are fused to improve the tracking accuracy of DeepSORT algorithm. The proposed algorithm is verified on the public data set MOT-16 and it’s tracking accuracy is enhanced to 4.1%.

Recently, spam emails have become a significant problem with the expanding usage of the Internet. It is to some extend obvious to filter emails. A spam filter is a system that detects undesired and malicious emails and blocks them from getting into the users' inboxes. Spam filters check emails for something "suspicious" in terms of text, email address, header, attachments, and language. However, we have used different features such as word2vec, word n-grams, character n-grams, and a combination of variable length n-grams for comparative analysis in our proposed approach. Different machine learning models such as support vector machine (SVM), decision tree (DT), logistic regression (LR), and multinomial naïve bayes (MNB) are applied to train the extracted features. We use different evaluation metrics such as precision, recall, f1-score, and accuracy to evaluate the experimental results. Among them, SVM provides 97.6 \% of accuracy, 98.8\% of precision, and 94.9\% of f1-score using a combination of n-gram features.

Since December 2019, the world is fighting against coronavirus disease (COVID-19). This disease is caused by a novel coronavirus termed as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). This work focuses on the applications of machine learning algorithms in the context of COVID-19. Firstly, regression analysis is performed to model the number of confirmed cases and death cases. Our experiments show that autoregressive integrated moving average (ARIMA) can reliably model the increase in the number of confirmed cases and can predict future cases. Secondly, a number of classifiers are used to predict whether a COVID-19 patient needs to be admitted to an intensive care unit (ICU) or semi-ICU. For this, classification algorithms are applied to a dataset having 5644 samples. Using this dataset, the most significant attributes are selected using features selection by ExtraTrees classifier, and Proteina C reativa (mg/dL) is found to be the highest-ranked feature. In our experiments, random forest, logistic regression, support vector machine, XGBoost, stacking and voting classifiers are applied to the top 10 selected attributes of the dataset. Results show that random forest and hard voting classifiers achieve the highest classification accuracy values near 98%, and the highest recall value of 98% in predicting the need for admission into ICU/semi ICU units.

Seismic imaging is the main technology used for subsurface hydrocarbon prospection. It~provides an image of the subsurface using the same principles as ultrasound medical imaging. As for any data acquired through hydrophones (pressure sensors) and/or geophones (velocity/acceleration sensors), the raw seismic data are heavily contaminated with noise and unwanted reflections that need to be removed before further processing. Therefore, the noise attenuation is done at an early stage and often while acquiring the data. Quality control (QC) is mandatory to give confidence in the denoising process and to ensure that a costly data re-acquisition is not needed. QC is done manually by humans and comprises a major portion of the cost of a typical seismic processing project. It is therefore advantageous to automate this process to improve cost and efficiency. Here, we propose a supervised learning approach to build an automatic QC system. The~QC system is an attribute-based classifier that is trained to classify three types of filtering (mild = under filtering, noise remaining in the data; optimal = good filtering; harsh = over filtering, the signal is distorted). The attributes are computed from the data and represent geophysical and statistical measures of the quality of the filtering. The system is tested on a full-scale survey (9000 km²) to QC the results of the swell noise attenuation process in marine seismic data.

Seismic imaging is the main technology used for subsurface hydrocarbon prospection. It provides an image of the subsurface using the same principles as ultrasound medical imaging. It is based on emitting a sound (pressure) wave through the subsurface and recording the reflected echoes using hydrophones (pressure sensors) and/or geophones (velocity/acceleration sensors). Contrary to medical imaging, which is done in real time, subsurface seismic imaging is an offline process that involves a huge volume of data and needs considerable computing power. The raw seismic data are heavily contaminated with noise and unwanted reflections that need to be removed before further processing. Therefore, the noise attenuation is done at an early stage and often while acquiring the data. Quality control (QC) is mandatory to give confidence in the denoising process and to ensure that a costly data re-acquisition is not needed. QC is done manually by humans and comprises a major portion of the cost of a typical seismic processing project. It is therefore advantageous to automate this process to improve cost and efficiency. Here, we propose a supervised learning approach to build an automatic QC system. The QC system is an attribute-based classifier that is trained to classify three types of filtering (mild = underfiltering, noise remaining in the data; optimal = good filtering; harsh = overfiltering, the signal is distorted). The attributes are computed from the data and represent geophysical and statistical measures of the quality of the filtering. The system is tested on a full-scale survey (9000 km2) to QC the results of the swell noise attenuation process in marine seismic data. The results are encouraging and helped identify localized issues that were difficult for a human to spot.

Significant progress has made in pattern recognition technology. However, one obstacle that has not yet overcome is the recognition of words in the Brahmi script, specifically the recognition of characters, compound characters, and word because of complex structure. For this kind of complex pattern recognition problem, it is always difficult to decide which feature extraction and classifier would be the best choice. Moreover, it is also true that different feature extraction and classifiers offer complementary information about the patterns to be classified. Therefore, combining feature extraction and classifiers, in an intelligent way, can be beneficial compared to using any single feature extraction. This study proposed the combination of HOG +zonal density with SVM to recognize the Brahmi words. Keeping these facts in mind, in this paper, information provided by structural and statistical based features are combined using SVM classifier for script recognition (word-level) purpose from the Brahmi words images. Brahmi word dataset contains 6,475 and 536 images of Brahmi words of 170 classes for the training and testing, respectively, and the database is made freely available. The word samples from the mentioned database are classified based on the confidence scores provided by support vector machine (SVM) classifier while HOG and zonal density use to extract the features of Brahmi words. Maximum accuracy suggested by system is 95.17% which is better than previously suggested studies.

Disease classification based on biological data is an important area in bioinformatics and biomedical research. It helps the doctors and medical practitioners for the early detection of disease and support them as a computer-aided diagnostic tool for accurate diagnosis, prognosis, and treatment of disease. Earlier Microarray gene expression data have wide application for the classification of disease, but now Next-generation sequencing (NGS) has replaced the Microarray technology. From the last few years, RNA sequence (RNA-Seq) data are widely used for the transcriptomic analysis. Hence, RNA-Seq based classification of disease is in its infancy. In this article, we present a general framework for the classification of disease constructed on RNA-Seq data. This framework will guide the researchers to process RNA-Seq, extract relevant features and apply the appropriate classifier to classify any kind of disease.

The electroencephalogram (EEG) is a record of brain activity. Brain Computer Interface (BCI) technology formed by the EEG signal has become one of the hotspots at present. How to extract the feature signal of EEG is the most basic research of BCI technology. In this paper, A new method of recognizing fatigue, conscious, concentrated state of human brain is proposed by the combination of discrete wavelet transform and the neural network based on EEG signal. First of all, the law signal is preprocessed by the wavelet denoising method because the law EEG signal contains a large number of high frequency noise, which is decomposed into multi-layer high frequency signal and low frequency signal. thus, δ wave, θ wave, α wave, β wave are obtained by the wavelet transform. And then, frequency band energy of the different wave is regards as the feature signal of EEG. In the experiment, the feature signal is classified by radial basic function (RBF) and annealed chaotic competitive learning network (ACCLN). RBF and ACCLN networks are trained with 500 sets of sample data and are tested by 100 sets of samples in different mental states. The experimental results show that the average accuracy of RBF network under three conditions are 88.75%, 88.25%, 88.5%, respectively, and the correct rate of ACCLN network is 97%, 98%, 98%, respectively.

Solar radiation prediction is an important process in ensuring optimal exploitation of solar energy power. Numerous models have been applied to this problem, such as numerical weather prediction models and artificial intelligence models. However, well-designed hybridization approaches that combine numerical models with artificial intelligence models to yield a more powerful model can provide a significant improvement in prediction accuracy. In this paper, we propose novel hybrid machine learning approaches that exploit auxiliary numerical data. The proposed hybrid methods invoke different machine learning paradigms, including feature selection, classification, and regression. Additionally, numerical weather prediction (NWP) models are used in the proposed hybrid models. Feature selection is used for feature space dimension reduction to reduce the large number of recorded parameters that affect estimation and prediction processes. The rough set theory is applied for attribute reduction and the dependency degree is used as a fitness function. We investigate the effect of the attribute reduction process with thirty different classification and prediction models in addition to the proposed hybrid model. Then, different machine learning models are constructed based on classification and regression techniques to predict solar radiation. Moreover, other hybrid prediction models are formulated to use the output of the numerical model of Weather Research and Forecasting (WRF) as learning elements in order to improve the prediction accuracy. The proposed methodologies are evaluated using a data set that is collected from different regions in Saudi Arabia.

The shift towards autonomous excavators in construction and mining is a significant leap towards enhancing operational efficiency and ensuring worker safety. However, it presents challenges such as the need for sophisticated decision-making and environmental perception due to complex terrains and diverse conditions. Our study introduces the E-GTN framework, a novel approach tailored for autonomous excavation that leverages advanced multisensor fusion and a custom-designed convolutional neural network to address these challenges. Results demonstrate that GridNet effectively processes grid data, enabling the reinforcement learning algorithm to make informed decisions, thereby ensuring efficient and intelligent autonomous excavator performance. The study concludes that the E-GTN framework offers a robust solution for the challenges in unmanned excavator operations, providing a valuable platform for future advancements in the field.

Object detection algorithms for open water aerial images present challenges such as small object size, unsatisfactory detection accuracy, numerous network parameters, and enormous computational demands. Current detection algorithms struggle to meet the accuracy and speed requirements while being deployable on small mobile devices. This paper proposes DFLM-YOLO, a lightweight small-object detection network based on the YOLOv8 algorithm with multiscale feature fusion. Firstly, to solve the class imbalance problem of the SeaDroneSee dataset, we propose a data augmentation algorithm called Small Object Multiplication (SOM). SOM enhance dataset balance by increasing the number of objects in specific categories, thereby improving model accuracy and generalization capabilities. Secondly, we optimize the backbone network structure by implementing Depthwise Separable Convolution (DSConv) and the newly designed FC-C2f, which reduces the model's parameters and inference time. Finally, we design the Lightweight Multiscale Feature Fusion Network (LMFN) to address the challenges of multiscale variations by gradually fusing the four feature layers extracted from the backbone network in three stages. In addition, LMFN incorporates the Dilated Re-param Block structure to increase the effective receptive field and improve the model's classification ability and detection accuracy. Experimental results on the SeaDroneSea dataset indicate that DFLM-YOLO improves mAP by 12.4% compared to the original YOLOv8s, while reducing parameters by 67.2%. This achievement provides a new solution for UAVs to conduct object detection missions in open water efficiently.

Proton Exchange Membrane Fuel Cell (PEMFC) is a fast-developing battery technology, and the key to its reliability and lifespan improvement lies in the accurate assessment of durability. However, the degradation mechanism of PEMFC is hard to determine and its internal parameters are highly coupled. Thus, the development of the more accurate life prediction model that meets the actual scenarios is urgent to investigate. To solve this problem, a multi-feature fusion life prediction method based on Temporal Convolutional Network-Gated Recurrent Unit (TCN-GRU) is proposed. A TCN algorithm is used as the prediction base model, and two GRU modules are included to the model to strengthen the model expression ability and improve its predictive accuracy. Two widely recognized datasets and two operating conditions are utilized for model training and prediction, respectively. Comparisons are made with single-feature parameter models in terms of Root Mean Square Error (RMSE) and Determination Coefficient (R2). The results show that the prediction accuracy of the TCN-GRU multi-feature fusion model is higher than that of the single-feature models in terms of stability and anti-interference under both operating conditions. Furthermore, with the increase of input feature parameters, the TCN-GRU model is closer to the real value, which proves once again that the proposed model can meet the accuracy requirements of the life prediction of PEMFC.

The fusion of global contextual information with local cropped block details is crucial for segmenting ultra-high resolution images. In this study, we introduce a novel fusion mechanism termed Global-Local Deep Fusion (GL-Deep Fusion) based on an enhanced transformer architecture, which efficiently integrates global contextual information and local details. Specifically, we propose the Global-Local Synthesis Networks (GLSNet), a dual-branch network where one branch processes the entire original image, while the other branch handles cropped local patches as input. The feature fusion of different branches in GLSNet is achieved through GL-Deep Fusion, significantly enhancing the accuracy of ultra-high resolution image segmentation. Particularly effective is GLSNet in identifying tiny overlapping items. To optimize GPU memory utilization, we meticulously design a dual-branch architecture that proficiently leverages the features it extracts, seamlessly integrating them into the enhanced transformer framework of GL-Deep Fusion. Extensive experiments conducted on challenging benchmarks, including DeepGlobe and Vaihingen datasets, demonstrate that GLSNet achieves a new state-of-the-art performance in terms of GPU memory utilization and segmentation accuracy tradeoff.

Throughout the evolution of machine learning, the size of models has steadily increased as researchers strive for higher accuracy by adding more layers. This escalation in model complexity necessitates enhanced hardware capabilities. Today, state-of-the-art machine learning models have become so large that effectively training them requires substantial hardware resources, which may be readily available to large companies but not to students or independent researchers. To make the research on machine learning models more accessible, this study introduces a size reduction technique that leverages stages in Pyramid Training and Similarity Comparison. Our results demonstrate that pyramid training can reduce model complexity while maintaining accuracy of conventional full-sized models, offering a scalable and resource-efficient solution for researchers and practitioners in hardware-constrained environments.

Despite UAV multi-target detection exhibits considerable developmental potential worldwide, it suffers distinct challenges compared with traditional tasks in this field. These challenges include insufficient feature extraction capabilities for small targets, limited capabilities for multi-dimensional feature fusion, as well as constraints on hardware computation parameters. Especially in mission scenarios such as disaster detection, these challenges will be further amplified. Consequently, this paper introduces SSE-YOLO, an innovative YOLO framework algorithm specially designed to address these challenges. To enhance the model's feature extraction capability, we employ the SPDConv module to replace the original Conv in the backbone section, utilizing depth-separable convolution instead of traditional convolution pooling. Concurrently, we eliminate the SPPF module at the bottom and address a new Separate Kernel Attention Pyramid Pooling (SKAPP) module, substantially enhancing the feature fusion capability at the model's core. Moreover, to address the challenge of multi-dimensional feature fusion, we replace the Concat module of the neck and head with E-BiFPN, transmitting feature information from the backbone to the lower network through four CBS blocks, which effectively resolves the issue of lost contextual information in the model. Meanwhile, SSE-YOLO undergoes ablation experiments on the VisDrone2019 dataset to evaluate its effectiveness against alternative methods, and experimental results illustrate the model's exceptional precision in detecting UAV targets. In comparison to models with comparable experimental accuracy, SSE-YOLO requires remarkably fewer parameters. On the VisDrone2019-DET-test-dev dataset, SSE-YOLO enhances mAP by 17.3%, with a 42.5% reduction in the parameter amounts. Therefore, the proposed method effectively tackles the challenge of reconciling low parameters and high accuracy, presenting a novel pathway for deep learning-based UAV multi-target detection.

To support pathologists in breast tumor diagnosis, deep learning plays a crucial role in the development of histological whole slide image (WSI) classification methods. However, automatic classification is challenging due to the high-resolution data and the scarcity of representative training data. To tackle these limitations, we propose a deep learning-based breast tumor gigapixel histological image multi-classifier integrated with a high-resolution data augmentation model to process the entire slide by exploring its local and global information and generating its different synthetic versions. The key idea is to perform classification and augmentation in feature latent space, reducing the computational cost while preserving the class label of the input. We adopt a deep learning-based multi-classification method and evaluate the contribution given by a Conditional Generative Adversarial Network-based data augmentation model on the classifier’s performance for three tumor classes in the BRIGHT Challenge dataset. The proposed method has allowed us to achieve an average F1 equal to 69.5, considering only the WSI dataset of the Challenge. The results are comparable to those obtained by the Challenge winning method (71.6), also trained on the annotated tumor region dataset of the Challenge.

Accurate weight measurement is pivotal for monitoring the growth and well-being of cattle. However, the conventional weighing process, which involves physically placing cattle on scales, is labor-intensive and distressing for the animals. Hence, the development of automated cattle weight prediction techniques assumes critical significance. This study proposes a weight prediction approach for Korean cattle using 3D segmentation-based feature extraction and regression machine learning techniques from incomplete 3D shapes acquired from real farm environments. In the initial phase, we generated mesh data of 3D Korean cattle shapes using a multiple-camera system. Subsequently, deep learning-based 3D segmentation with the PointNet network model was employed to segment two dominant parts of the cattle. From these segmented parts, three crucial dimensions of Korean cattle were extracted. Finally, we implemented five regression machine learning models (CatBoost regression, LightGBM, Polynomial regression, Random Forest regression, and XGBoost regression) for weight prediction. To validate our approach, we captured 270 Korean cattle in various poses, totaling 1190 poses of 270 cattle. The best result was achieved with mean absolute error (MAE) of 25.2 kg and mean absolute percent error (MAPE) of 5.81% using the random forest regression model.

Machine Learning techniques can be used to identify whether deficits in cognitive functions contribute to antisocial and aggressive behavior. This paper initially presents the results of tests conducted on delinquent and non-delinquent youths to assess their cognitive functions. The dataset extracted from these assessments, consisting of 37 predictor variables and one target. was used to train three algorithms that aim to predict whether the data corresponds to that of a young offender or a non-offending youth. Prior to this, statistical tests were conducted on the data to identify characteristics that exhibited significant differences in order to select the most relevant features and optimize the prediction results. Additionally, other feature selection methods, such as Boruta, RFE, and Filter, were applied, and their effects on the accuracy of each of the three machine learning models used (SVM, RF, and KNN) were compared. 80% of the data were utilized for training, while the remaining 20% were used for validation. The best result was achieved by the K-NN model trained with 19 features selected by the Boruta method, followed by the SVM model trained with 24 features selected by the filter method.

Covid-19 is a lung disease caused by a Coronavirus family virus. Due to its extraordinary prevalence and death rates, it has spread quickly to every country in the world. Thus, achieving peaks and outlines and curing different types of relapses is extremely important. Given the worldwide prevalence of Coronavirus and the participation of physicians in all countries, Information has been gathered regarding the properties of the virus, its diverse types, and the means of analyzing it. Numerous approaches have been used to identify this evolving virus. It is generally considered the most accurate and acceptable method of examining the patient's lungs and chest through a CT scan. As part of the feature extraction process, a method known as fractional Fourier transform (FrFT) has been applied as one of the time-frequency domain transformations. The proposed method was applied to a database consisting of 2481 CT images. Following the transformation of all images into equal sizes and the removal of non-lung areas, multiple combination windows are used to reduce the number of features extracted from the images. In this paper, the results obtained for KNN and SVM classification have been obtained with accuracy values of 99.84% and 99.90%, respectively.

Purpose: Emotional voice conveys important social cues that demand listeners’ attention and timely processing. This event-related potential study investigated the feasibility of a multi-feature oddball paradigm to examine adult listeners’ neural responses to detecting emotional prosody changes in non-repeating naturally spoken words. Method: Thirty-three adult listeners completed the experiment by passively listening to the words in neutral and three alternating emotions while watching a silent movie. Previous research documented pre-attentive change-detection electrophysiological responses (e.g., MMN, P3a) to emotions carried by fixed syllables or words. Given that the MMN and P3a have also been shown to reflect extraction of abstract regularities over repetitive acoustic patterns, the current study employed a multi-feature oddball paradigm to compare listeners’ MMN and P3a to emotional prosody change from neutral to angry, happy, and sad emotions delivered with hundreds of non-repeating words in a single recording session. Results: Both MMN and P3a were successfully elicited by the emotional prosodic change over the varying linguistic context. Angry prosody elicited the strongest MMN compared to happy and sad prosodies. Happy prosody elicited the strongest P3a in the centro-frontal electrodes, and angry prosody elicited the smallest P3a. Conclusions: The results demonstrated that listeners were able to extract the acoustic patterns for each emotional prosody category over constantly changing spoken words. The findings confirm the feasibility of the multi-feature oddball paradigm in investigating emotional speech processing beyond simple acoustic change detection, which may potentially be applied to pediatric and clinical populations.

A major problem in the post-inversion geophysical interpretation is the extraction of geological information from inverted physical property models, which do not necessarily represent all underlying geological features. No matter how accurate the inversions are, each inverted physical property model is sensitive to limited aspects of subsurface geology and is insensitive to other geological features that are otherwise detectable with complementary physical property models. Therefore, specific parts of the geological model can be reconstructed from different physical property models. To show how this reconstruction works, we simulated a complex geological system that comprises an original layered earth model that has passed several geological deformations and alteration overprints. Linear combination of complex geological features comprised three physical property distributions: Electrical resistivity, induced polarization chargeability, and magnetic susceptibility models. This study proposes a multivariate feature extraction approach to extract information about the underlying geological features comprising the bulk physical properties. We evaluated our method in numerical simulations and compared three feature extraction algorithms to see the tolerance of each method to the geological artifacts and noises. We show that the fast-independent component analysis (fast-ICA) algorithm by negentropy maximization is a robust method in the geological feature extraction that can handle the added unknown geological noises. The post-inversion physical properties are also used to reconstruct the underlying geological sources. We show that the sharpness of the inverted images is an important constraint on the feature extraction process. Our method successfully separates geological features in multiple 3D physical property models. This methodology is reproducible for any number of lithologies and physical property combinations and can recover the latent geological features, including the background geological patterns from overprints of chemical alteration.

State-of-the-art remote sensing scene classification methods employ different Convolutional Neural Network architectures for achieving very high classification performance. A trait shared by the majority of these methods is that the class associated with each example is ascertained by examining the activations of the last fully connected layer, and the networks are trained to minimize the cross-entropy between predictions extracted from this layer and ground-truth annotations. In this work, we extend this paradigm by introducing an additional output branch which maps the inputs to low dimensional representations, effectively extracting additional feature representations of the inputs. The proposed model imposes additional distance constrains on these representations with respect to identified class representatives, in addition to the traditional categorical cross-entropy between predictions and ground-truth. By extending the typical cross-entropy loss function with a distance learning function, our proposed approach achieves significant gains across a wide set of benchmark datasets in terms of classification, while providing additional evidence related to class membership and classification confidence.

We present a method to improve the reconstruction and generation performance of variational autoencoder (VAE) by injecting an adversarial learning. On the other hand, instead of comparing the reconstructed with the original data to calculate the reconstruction loss, we use a consistency principle for deep features. The training process of the VAE is then divided into two steps, training the encoder and then training the decoder. By using this two-step learning process, our method can be more widely used in applications other than image processing. While training the encoder, the label information is integrated to better structure the latent space in a supervised way. The adversarial constraints allow the decoder to generate data with better authenticity and more realistic than the conventional VAE. We present experimental results to show that our method gives better performance than the original VAE.

Since the Coronavirus disease (COVID-19) outbreak at the end of 2019, the past two month has seen an acceleration both in and outside China in the R&D of the diagnostics, vaccines and therapeutics for this novel coronavirus. As one of the molecular forces that determine protein structure, electrostatic effects dominate many aspects of protein behaviour and biological function. Thus, incorporating currently available experimental structures related to COVID-19, this article reports a simple python-based analysis tool and a LaTeX-based editing tool to extract and summarize the electrostatic features from experimentally determined structures, to strengthen our understanding of COVID-19's structure and function and to facilitate machine-learning and structure-based computational design of its neutralizing antibodies and/or small molecule(s) as potential therapeutic candidates. Finally, this article puts forward a brief update of the structurally observed electrostatic features of the COVID-19 coronavirus.

In order to realize rapid and nondestructive monitoring of wheat biomass in field, field experiments based on different densities, nitrogen fertilizer and variety treatments were studied. RGB images of wheat in the main growth stage were obtained by UAV, and wheat color and texture feature indices were obtained by image processing, and wheat biomass was obtained by field sampling in the same period. Then the relationship between different color and texture feature indices and wheat biomass was analyzed to select the color and texture feature index suitable for wheat biomass estimation. The results showed that there was a high correlation between image color index and wheat biomass in different stages, and most of them reached a very significant correlation level. However, the correlation between image texture feature index and wheat biomass was poor, only a few indexes reached significant or extremely significant correlation level. Based on the above results, the color indices with the highest correlation to wheat biomass or the combining indices of color and texture feature in different growth stage were used to construct estimation model of wheat biomass. The models were validated using independently measured biomass data, and the correlation between simulated and measured values reached the significant level, RMSE were smaller. This indicated that the estimated results by the models were reliable and accurate. It also showed that the estimation models of wheat biomass combined with color and texture feature indices of UAV image were better than the single color index models. The results would provide a new method for real-time monitoring of wheat field growth and biomass estimation.

Many corner detector techniques have already been used in extracting information from UAV images to perform various photogrammetric and mapping activities. Among these techniques is the Feature from Accelerated Segment Test (FAST) and the Harris corner detector. It is widely agreed that the evaluation of detectors is of great importance because it evaluates and enhances the accuracy of the detected features. This research evaluates the performance of FAST-9 and FAST-12 as well as the Harris detector in terms of the repeatability rate, completeness, and correctness under different threshold values. Each method is evaluated in terms of its ability for detection UAV objects (crowd and cars features). Then the common detected features between both FAST versions and the Harris detector are extracted. This is to determine which method performs best under different image conditions (e.g., illumination variations, camera position and orientation, and image noise). The results show that the size of the threshold plays a crucial role in determining the number of detected feature points. An increase in the threshold value leads to a decrease in the number of detected points and vice versa. Thus, the correctness decreases whereas the completeness increases as a function of the threshold values. Furthermore, the relationship between the FAST-9 and the Harris detector is slightly better than those between the FAST-12 and the Harris detector. This is because the number of common features between the FAST-9 and the Harris detector are relatively higher than those between the FAST-12 and the Harris detector.

The objective to fast-track the mapping and registration of large numbers of unrecorded land rights globally, leads to the experimental application of Artificial Intelligence (AI) in the domain of land administration, and specifically the application of automated visual cognition techniques for cadastral mapping tasks. In this research, we applied and compared the ability of rule-based systems within Object Based Image Analysis (OBIA), as opposed to human analysis, to extract visible cadastral boundaries from Very high resolution (VHR) World View-2 image, in both rural and urban settings. From our experiments, machine-based techniques were able to automatically delineate a good proportion of rural parcels with explicit polygons where the correctness of the automatically extracted boundaries was 47.4% against 74.24% for humans and the completeness of 45% for machine, as against 70.4% for humans. On the contrary, in the urban area, automatic results were counterintuitive: even though urban plots and buildings are clearly marked with visible features such as fences, roads and tacitly perceptible to eyes, automation resulted in geometrically and topologically poorly structured data, that could neither be geometrically compared with human digitised, nor actual cadastral data from the field. These results provide an updated snapshot with regards to the performance of contemporary machine-drive feature extraction techniques compared to conventional manual digitising.

Edge detection is one of the key issues in the field of computer vision and remote sensing image analysis. Although many different edge-detection methods have been proposed for gray-scale, color, and multispectral images, they still face difficulties when extracting edge features from hyperspectral images (HSIs) that contain a large number of bands with very narrow gap in the spectral domain. Inspired by the clustering characteristic of the gravitation, a novel edge-detection algorithm for HSIs is presented in this paper. In the proposed method, we first construct a joint feature space by combining the spatial and spectral features. Each pixel of HSI is assumed to be a celestial object in the joint feature space, which exerts gravitational force to each of its neighboring pixel. Accordingly, each object travels in the joint feature space until it reaches a stable equilibrium. At the equilibrium, the image is smoothed and the edges are enhanced, where the edge pixels can be easily distinguished by calculating the gravitational potential energy. The proposed edge-detection method is tested on several benchmark HSIs and the obtained results were compared with those of three state-of-the-art approaches. The experimental results confirm the efficacy of the proposed method

Alzheimer's disease (AD) detection acting as an essential role in global health care due to misdiagnosis and sharing many clinical sets with other types of dementia, and costly monitoring the progression of the disease over time by magnetic reasoning imaging (MRI) with consideration of human error in manual reading. This paper goal a comparative study on the performance of data mining techniques on two datasets of Clinical and Neuroimaging Tests with AD. Our proposed model in the first stage, Apply clinical medical dataset to a composite hybrid feature selection (CHFS), for extract new features to select the best features due to eliminating obscures features, In parallel with Apply a novel hybrid feature extraction of three batch edge detection algorithm and texture from MRI images dataset and optimized with fuzzy 64-bin histogram. In the second stage, we applied a clinical dataset to a stacked hybrid classification(SHC) model to combine Jrip and random forest classifiers with six model evaluations as meta-classifier individually to improve the prediction of clinical diagnosis. At the same stage of improving the classification accuracy of neuroimaging (MRI) dataset images by applying a convolution neural network (CNN) in comparison with traditional classifiers, running on extracted features from images. The authors have collected the clinical dataset of 426 subjects with (1229 potential patient sample) from oasis.org and (MRI) dataset from a benchmark kaggle.com with a total of around ~5000 images each segregated into the severity of Alzheimer's. The datasets evaluated using an explorer set of weka data mining software for the analysis purpose. The experimental show that the proposed model of ‏(CHFS) feature extraction ‏ lead to effectively reduced the false-negative rate with a relatively high overall accuracy with a stack hybrid classification of support vector machine (SVM) as meta-classifier of 96.50% compared to 68.83% of the previous result on a clinical dataset, Besides a compared model of CNN classification on MRI images dataset of 80.21%. The results showed the superiority of our CHFS model in predicting Alzheimer's disease more accurately with the clinical medical dataset in early-stage compared with the neuroimaging (MRI) dataset. The results of the proposed model were able to predict with accurately classify Alzheimer's clinical samples at a low cost in comparison with the MRI-CNN images model at the early stage and get a good indicator for high classification rate for MRI images when applying our proposed model of SHC.

The study offers an in-depth investigation of the impact of climate conditions on various terrain features in the Kaffiøyra area, southwest Svalbard, focusing on the changes of shorelines, glacier termini, and outwash areas. We utilized cost-effective, consistent remote sensing data and GIS techniques to analyze historical long-term changes in these terrain features, evaluated the effects of temperature variations on glacier retreat, and developed a predictive model for future glacier area changes based on temperature data from climate change scenarios and the ground observation station near the site. The results demonstrated a range of responses in shoreline stability and changes across different zones in the study site. Based on the predefined baseline, the glacier-shorelines (tidewater glacier) exhibit significant erosion, with the averaged rates varying from -64.7 to + 9.2 m/yr. However, the land-shorelines show relatively stable, with changes ranging from -2.2 to +3.7 m/yr, and the outwash areas exhibited minor increases of less than 10% as compared to the data obtained in 1985. The analysis of seven glacier termini indicated a general reduction in the glacier area; notably, Aavatsmarkbreen, a tidewater glacier, retained only 32% of its original size by 2023. The land glaciers such as Waldemarbreen, Irenebreen, Elisebreen, Eivindbreen, Andreasbreen, and Oliverbreen, preserve 54.8%, 61.3%, 65.0%, 74.0%, 55.5%, and 43.7% of their areas, respectively. A marked negative linear correlation between temperature and remaining glacier areas was observed, enabling the prediction of future glacier variations under various climate conditions. The study outlines the evolution of terrain features in Kaffiøyra in nearly 40 years. Results show the relative stability of land-shorelines, significant glacier retreats in response to warming temperatures, and slight growth in outwash areas. The study highlights the critical need for robust land monitoring systems that enhance our understanding and enable effective decision-making through predictive modeling of climate impacts.

Recommending products effectively based on historical transaction data and metadata is crucial for improving user satisfaction and increasing sales. This study presents a recommendation system evaluated using Mean Average Precision (MAP@12), leveraging a diverse dataset that includes apparel types, customer demographics, product descriptions, and images. Our approach integrates Mult-Recall methods and LightGBM Ranker, utilizing various recall strategies and advanced feature engineering techniques to enhance recommendation accuracy. By combining these advanced machine learning models and ensemble techniques, our proposed solution outperforms existing methods, providing a robust and efficient recommendation system tailored to diverse customer needs and product characteristics. This integrated approach not only addresses the challenges of data sparsity and cold start but also adapts to the dynamic nature of user preferences and market trends. The results demonstrate significant improvements in both recommendation precision and user satisfaction, highlighting the potential of our approach for real-world e-commerce applications.

Classifiers face a myriad of challenges in today’s data-driven world, ranging from overfitting and high computational costs to low accuracy, imbalanced training datasets, and the notorious black box effect. Furthermore, many traditional classifiers struggle with the robust handling of noisy and missing feature values. In response to these hurdles, we present classification methods that leverage the power of feature partitioning learning and outranking measures. Our classification algorithms offer an innovative approach, eliminating the need for prior domain knowledge by automatically discerning feature intervals directly from the data. These intervals capture essential patterns and characteristics within the dataset, empowering our classifiers with newfound adaptability and insight. In addition, we employ outranking measures to mitigate the influence of noise and uncertainty in the data. Through pairwise comparisons of alternatives on each feature, we enhance the robustness and reliability of our classification outcomes. The developed classifiers are empirically evaluated on several data sets from UCI repository and are compared with well-known classifiers including k Nearest Neighbors (K-NN), Support Vector Machine (SVM), Random Forest (RF), Neural Network (NN), Naive Bayesian (NB) and Nearest Centroid (NC). The experiments result demonstrate that the classifiers based on feature interval learning and outranking approaches are robust to imbalanced data and to irrelevant features and achieve comparably and even better performances than the well-known classifiers in some cases. Moreover, our proposed classifiers produce more explainable models whilst preserving high predictive performance levels.

Aiming at the problems of inaccurate segmentation of long object and information loss of small object in real-time semantic segmentation algorithm, this paper proposes a lightweight multi-branch real-time semantic segmentation network based on BiseNetV2. The new auxiliary branch makes full use of spatial details and context information to cover the long object in the field of view. Meanwhile, in order to ensure the inference speed of the model, the asymmetric convolution is used in each stage of the auxiliary branch to design a structure with low computational complexity. In the multi-branch fusion stage, the alignment and fusion module is designed to provide guidance information for deep and shallow feature mapping, so as to make up for the problem of feature misalignment in the fusion of information at different scales, and thus reduce the loss of small target information. In order to further improve the model’s awareness of key information, a global context module is designed to capture the most important features in the input data. The proposed network uses NVIDIA GeForce RTX 3080 Laptop GPU experiment, on the road street view data set Cityscapes and CamVid average occurring simultaneously ratio reached 77.1% and 78.4% respectively, with running speed of 127 frames/s respectively and 112 frames/s. The experimental results show that the proposed algorithm can achieve real-time segmentation and improve the accuracy significantly, showing good semantic segmentation performance

In this paper, we present a method for estimating GPS signals from visual information captured by a monocular camera mounted on a fixed-wing Unmanned Aerial Vehicle. The main challenge in visual odometry from aerial images is the computation of scale due to irregularities in the elevation of the terrain. That is, it is not possible to accurately convert from pixels in the image to meters in space, and the error accumulates. The contribution of this work is to reduce the accumulated error by comparing the images from the camera with satellite images. The algorithm has been tested in real-time experiments, and the results show that it is possible to eliminate the accumulated error using satellite images.

Garlic, as an important economic crop in China, still has room for improvement in terms of identification using remote sensing technology, Among them, high-precision classification of garlic has become an important subject. The Erhai Lake is an important freshwater lake in China. Under the influence of technology and policies, significant changes have occurred in the cultivation of garlic crops. this study constructed multidimensional features for crop classification suitable for Google Earth Engine, and Propose a method for identifying garlic crops using sample and feature datasets under limited conditions. The results indicate that: 1) In the land-use classification of the Erhai Lake Basin, the importance ranking of characteristic bands, from high to low, is as follows: spectral features, vegetation features, texture features, and terrain features. 2) The Random Forest method based on feature selection demonstrates high classification accuracy in land-use classification within the Erhai Lake Basin in Yunnan Province. The overall classification accuracy reaches 95.79%, with a Kappa coefficient of 0.9481. 3) The expansion direction of garlic cultivation in the Erhai Lake Basin initially strengthened and then weakened from 1999 to 2023. The vertical development of garlic cultivation reached saturation, showing a slow trend towards horizontal expansion between 2005 and 2018. The planting distribution in various townships in the Erhai Lake Basin gradually shifted from a relatively uniform distribution to an upstream development in the basin. This study utilizes the Google Earth Engine (GEE) cloud computing platform and machine learning algorithms to compensate for the lack of statistical data on garlic cultivation in the Erhai Lake Basin. Simultaneously, it accurately, rapidly, and efficiently extracts planting information, demonstrating significant potential for practical applications.

Passive marine listening, involving the acoustic monitoring of underwater environments, has become increasingly vital for scientific research, environmental monitoring, and defense applications. However, the success of such applications critically depends on the ability to extract meaningful information from the often noisy and dynamic underwater acoustic environment. In this context, Mel-frequency cepstral coefficients (MFCCs) have emerged as an essential tool for feature extraction, capturing the spectral characteristics of marine sounds and making them ideal for species identification and sound event detection. This research presents an innovative approach to enhance the robustness of MFCC-based passive marine listening through adaptive noise reduction techniques. The proposed approach utilizes dynamic spectral subtraction to counter underwater noise, resulting in enhanced signal-to-noise ratios for desired sounds. This adaptive system adjusts to dynamic underwater noise, allowing it to concentrate on target sounds and suppress interference effectively. The experimentation further validates the effectiveness of the proposed approach, with results reaching 99% on the full dataset of 570 mammals and vessels, demonstrating the effectiveness of SVM and Random Forests in the classification of underwater audio.

Aiming at the classification identification problem of aeroengines, this paper adopts telemetry Fourier Transform Infrared Spectrometer to collect aeroengine hot jet infrared spectrum data, and proposes an aeroengine classification identification method based on spectral feature vectors. First, aeroengine hot jet infrared spectrum data are acquired and measured, meanwhile, the spectral feature vectors based on CO2 are constructed. Then, the feature vectors are combined with the seven mainstream classification algorithms to complete the training and prediction of the classification model. In the experiment, two Fourier transform infrared spectrometers, EM27 developed by Bruker and a self-developed telemetry FTIR, were used to telemeter the hot jet of three aeroengines to obtain infrared spectral data. The training data set and test data set were randomly divided in a ratio of 3:1. The model training of the training data set and the label prediction of the test data set were carried out by combining spectral feature vectors and classification algorithms. The classification evaluation indicators were accuracy, precision, recall, confusion matrix, and F1-score. The classification recognition accuracy of the algorithm was approximately 98%. This paper has great significance for the fault diagnosis of aeroengines and classification recognition of aircrafts.

In the intricate domain of text analysis, syntactic dependency trees play a crucial role in unraveling the web of relations among entities embedded in the text. These trees provide a structural roadmap, guiding us through the complex syntax to pinpoint the interactions and connections between different entities. However, the challenge lies in sifting through this intricate structure to extract relevant information, a task that requires precision and discernment. Traditional approaches often rely on rule-based pruning methods to simplify these dependency structures, focusing on certain parts while discarding others. Yet, this approach has its pitfalls, as it can overlook critical nuances and connections that are vital for a comprehensive understanding of the text. Addressing this gap, our research introduces the Syntactic Dependency-Aware Neural Networks (SDANNs), a groundbreaking model designed to harness the full power of the entire dependency tree. This approach marks a significant departure from the conventional methods. Instead of the rigid rule-based pruning, SDANNs implement a more flexible and dynamic 'soft-pruning' technique. This method allows the model to adaptively focus on the sub-structures within the dependency tree that are most relevant for understanding the relationships between entities. By doing so, it ensures that no vital information is overlooked, and all potential connections are considered. The efficacy of SDANNs is not just theoretical but has been empirically validated through extensive testing and evaluations across a wide range of tasks. These tasks include the extraction of complex relations spanning multiple sentences, as well as detailed analyses at the sentence level. In each of these scenarios, SDANNs have demonstrated a remarkable ability to leverage the full structural complexity of dependency trees. This capability sets them apart from existing models, enabling a more nuanced and comprehensive analysis of textual relations. The results of these evaluations consistently show that SDANNs not only meet but significantly exceed the performance of prior models. This superiority is evident in the way SDANNs handle the multifaceted and often subtle interactions within the text, offering insights that were previously inaccessible with conventional methods. In summary, the Syntactic Dependency-Aware Neural Networks represent a significant advancement in the field of text analysis. By fully embracing the complexity of syntactic dependency trees and employing a sophisticated 'soft-pruning' approach, SDANNs open new avenues for exploring and understanding the intricate relationships that exist within written language. This model stands as a testament to the potential of combining advanced neural network architectures with a deep understanding of linguistic structures, paving the way for more accurate, nuanced, and comprehensive analyses of text.

An algorithm is proposed for discriminating the fatigue state of air traffic controllers based on ap-plying multispeech feature fusion using an FSVM to voice data, and for extracting eye-fatigue-state discrimination features based on PERCLOS eye data. For the speech algorithm and an eye-fatigue index, a new controller fatigue-state evaluation index based on the entropy weight method is proposed based on decision-level fusion of fatigue discrimination results for speech and the eyes. Experimental results show that the fatigue-state recognition accuracy rate was 84.81% for the fatigue state evaluation index, which was 3.36% and 1.86% higher than those for speech and eye assessments, respectively. The comprehensive fatigue evaluation index provides important reference values for controller scheduling and mental-state evaluations.

Machinery degradation assessment can offer meaningful prognosis and health management in-formation. Although numerous machine prediction models based on artificial intelligence have emerged in recent years, they still face a series of challenges: (1) Many models continue to rely on manual feature extraction. (2) Deep learning models still struggle with long sequence prediction tasks. (3) Health indicators are inefficient for remaining useful life (RUL) prediction with cross-operational environments when dealing with high-dimensional datasets as inputs. This research proposes a health indicator construction methodology based on a transformer self-attention transfer network (TSTN). This methodology can directly deal with the high-dimensional raw dataset and keep all the information without missing when the signals are taken as the input of the diagnosis and prognosis model. First, we design an encoder with a long-term and short-term self-attention mechanism to capture crucial time-varying information from a high-dimensional dataset. Second, we propose an estimator that can map the embedding from the encoder output to the estimated degradation trends. Then, we present a domain dis-criminator to extract invariant features from different machine operating conditions. The case studies with the FEMTO-ST bearing dataset and the Monte Carlo method for RUL prediction during the degradation process are conducted. The experiment results fully exhibited the signif-icant advantages of the proposed method compared to other state-of-the-art techniques.

Human Action Recognition is widely used in multiple fields to recognize activities and extract spatial and temporal information. This paper focuses on analyzing various methods as well as provides extensive knowledge of the foundational concepts of the HAR. A dataset is also very crucial for any research study so we discussed the popular dataset and its features in this paper.

Deep networks-based hashing has gained significant popularity in recent years, particularly in the field of image retrieval. However, most existing methods only focus on extracting semantic information from the final layer, disregarding valuable structural information that contains important semantic details crucial for effective hash learning. To address this limitation and improve image retrieval accuracy, we propose a novel deep hashing method called Deep Supervised Hashing by Fusing Multiscale Deep Features (DSHFMDF). Our approach involves extracting multiscale features from multiple convolutional layers and fusing them to generate more robust representations for efficient image retrieval. Experimental results on CIFAR10 and NUS-WIDE datasets demonstrate that our method surpasses the performance of state-of-the-art hashing techniques.

To address the deficiency of existing relation extraction models in effectively extracting relational triplets pertaining to railway traffic knowledge in Tibet, this paper constructs a Tibet Railway Traffic text dataset and provides an enhanced relation extraction model. The proposed model incorporates subject feature enhancement and relational attention mechanisms. It leverages a pre-trained model as the embedding layer to obtain vector representations of text. Subsequently, the subject is extracted and its semantic information is augmented using an LSTM neural network. Furthermore, during object extraction, the multi-head attention mechanism enables the model to prioritize relations associated with the aforementioned features. Finally, objects are extracted based on the subjects and relations. The proposed method has been comprehensively evaluated on multiple datasets, including the Tibet Railway Traffic text dataset and two public datasets. The results on the Tibet dataset achieves an F1-score of 93.3\%, surpassing the baseline model CasRel by 0.8\%, indicating a superior applicability of the proposed model. On the other hand, the model achieves F1-scores of 91.1\% and 92.6\% on two public datasets, NYT and WebNLG, respectively, outperforming the baseline CasRel by 1.5\% and 0.8\%, which highlights the good generalization ability of the proposed model.

This paper explores the use of different deep learning models for predicting precipitation in 56 meteorological stations in Jilin Province, China. The models used include Stacked-LSTM, Transformer, and SVR, and Gaussian noise is added to the data to improve their robustness. Results show that the Stacked-LSTM model performs the best, achieving high prediction accuracy and stability. The study also conducts variable attribution analysis using LightGBM and finds that temperature, dew point, precipitation in previous days, and air pressure are the most important factors affecting precipitation prediction, which is consistent with traditional meteorological theory. The paper provides detailed information on the data processing, model training, and parameter settings, which can serve as a reference for future precipitation prediction tasks. The findings suggest that adding Gaussian noise to the dataset can improve the model's generalization ability, especially for predicting days with zero precipitation. Overall, this study provides useful insights into the application of deep learning models in precipitation prediction and can contribute to the development of meteorological forecasting and applications.

Taste determination in small molecules is critical in food chemistry, but traditional experimental methods can be time-consuming. Consequently, computational techniques have emerged as val-uable tools for this task. In this study, we explore taste prediction using various molecular feature representations and assess the performance of different machine learning algorithms on a dataset comprising 2,601 molecules. The results reveal that GNN-based models outperform other ap-proaches in taste prediction. Moreover, consensus models that combine diverse molecular repre-sentations demonstrate improved performance. Among these, molecular fingerprints + GNN con-sensus model emerges as the top performer, highlighting the complementary strengths of GNNs and molecular fingerprints. These findings have significant implications for food chemistry research and related fields. By leveraging these computational approaches, taste prediction can be expedited, leading to advancements in understanding the relationship between molecular structure and taste perception in various food components and related compounds.

Due to the influence of background conditions, lighting conditions, occlusion issues and the image resolution, how to extract robust person features is one of the difficulties in ReID research. Vision in Transformers (ViT) has achieved significant results in the field of computer vision. However, the existing problems still limit its application in ReID due to slow extraction of person features and difficulty in utilizing local features of people. To solve the mentioned problems, we utilize Pyramid Vision Transformer (PVT) as the backbone of feature extraction and propose a PVT-based ReID method in conjunction with other studies. Firstly, some improvements suitable for ReID are used on the PVT backbone, and we establish a basic model by using powerful methods verified on CNN-based ReID. Secondly, in an effort to further promote the robustness of the person features extracted by the PVT backbone, two new modules are designed. (1) The local feature clustering (LFC) is recommend to enhance the robustness of person features by calculating the distance between local features and global feature to select the most discrete local features and clustering them. (2) The side information embeddings (SIE) are used to encode non-visual information and send it into the network for training to reduce its impact on person features. Finally, the experiments show that PVTReID has achieved excellent results in ReID datasets and are 20% faster on average than CNN-based ReID methods.

Deep convolutional neural networks (CNN) are favored methods widely used in medical image processing due to their assured shown performance. Recently, the emergence of new lung diseases and the possibility of early detection of their symptoms has attracted many researchers to classify diseases by training deep CNNs on lung CT images. The trained networks are expected to distinguish between lung indications in different diseases, especially at the early stages of them. With the hope of achieving this purpose, we proposed an efficient deep CNN called AFEX-Net with adaptive feature extraction layers that successfully extract distinguishing features and classify chest CT images. The efficiency of the proposed network has two aspects: it is a lightweight network with low number of parameters and fast training and it has adaptive pooling layers and adaptive activation functions to increase its level of compatibility to the input data. The proposed network has been evaluated on a dataset with more than 10K chest CT slices, while an efficient pre-processing method is developed to remove any bias from the images. Additionally, we evaluated the performance of the proposed model on the public COVID-CTset dataset to prove the generalisability of our model. The obtained results confirm the competence of the proposed network in confronting medical images, where prompt and accurate learning is required.

With the growing popularity of smartphones, user identification has become an essential component of maintaining security and privacy. This study investigates how smartphone accelerometer data can be used to identify users, and it makes recommendations for the ideal application parts. Accelerometer data from the HMOG public dataset was used to train deep learning, conventional classifiers, and voting classifiers, which were then utilized to identify users. To enhance performance, feature selection and pre-processing techniques were researched. The results show that RFE feature selection outperforms other approaches and that LSTM followed by XGBoost has the best identification performance as indicated by a relatively large number of machine learning performance measures. The proposed identification system nevertheless performed well and outperformed existing methods, which were principally created and tested on the same HMOG public smartphone dataset, even with a larger number of users. Further work would be necessary for such an application to reach its full potential, though.

Cross flow heat exchangers are commonly used in the thermal industry to transfer heat from hot tubes to cooling fluid. To protect the heat exchanger tubes from corrosion and dust accumulation, microscale coatings are often applied. In this study, we present machine-learning models for predicting heat transfer from hot tubes with different micro-sized coatings to cooling fluid in a turbulent flow using computational fluid dynamics simulations. A dataset of approximately 1000 cases was generated by varying the coating coverage thickness of each tube, the inlet Reynolds number, fluid flow inlet temperature, and wall temperature of tubes. The machine-learning models were generated to predict the overall heat flow rate in the heat exchanger, and it was found that combining the features based on their importance preserved the accuracy of the models while maintaining all the relevant information. The simulation results demonstrate that the proposed method increases the coefficient of determination (R2) for the models. The R2 values for unseen data for Random Forest, K-Nearest Neighbors, and Support Vector Regression were 0.9810, 0.9037, and 0.9754, respectively, indicating the usefulness of the proposed model for predicting heat transfer in various types of heat exchangers.

This research investigates the potential benefits of integrating machine learning algorithms and IoT sensors in modern agriculture. The focus is on optimizing crop production and reducing waste through informed decisions about planting, watering, and harvesting crops. The paper discusses the current state of machine learning and IoT in agriculture, highlighting key challenges and opportunities. It also presents experimental results that demonstrate the impact of changing labels on the accuracy of data analysis algorithms. The findings recommend that by analyzing wide-ranging data collected from farms, including real-time data from IoT sensors, farmers can make more informed verdicts about factors that affect crop growth. Eventually, the integration of these technologies can transform modern agriculture by increasing crop yields while minimizing waste. In our studies, we achieve a classification accuracy of 99.59% using the Bayes Net algorithm and 99. 46% using Naïve Bayes Classifier, and Hoeffding Tree algorithms. Our results indicate that we achieved high accuracy results in our experiments in order to increase crop growth.

Convolutional neural networks (CNN) have attracted much attention as a commonly used method for hyperspectral image (HSI) classification in recent years, however, CNNs can only be applied to Euclidean data and have limitations in dealing with relationships due to the limitations of local feature extraction. However, each pixel of a hyperspectral image contains a set of spectral bands that are correlated and interact with each other, and the methods used to process Euclidean data cannot effectively obtain these correlations. In contrast, the graph convolutional network (GCN) can be used in non-Euclidean data, but usually leads to oversmoothing and ignoring local detail features due to the need for superpixel segmentation processing to reduce computational effort. To overcome the above problems, we constructed a network a fusion network based on GCN and CNN, which contains two branches: a graph convolutional network based on superpixel segmentation and a convolutional network with added attention mechanism. The graph convolu-tional branch can extract the structural features and capture the relationships between the nodes, and the convolutional branch can extract the detailed features in the local fine region. Owing to the fact that the features extracted from the two branches are different, the classification performance can be improved by fusing the complementary features extracted from the two branches. To vali-date the proposed algorithm, experiments were conducted on three widely used datasets, namely Indian Pines, Pavia University, and Salinas, and the overall accuracy of 98.78% was obtained in the Indian Pines dataset, and the overall accuracy of 98.99% and 98.69% was obtained in the other two datasets. The results showed that the proposed fusion network can obtain richer features and achieve high classification accuracy.

Lactobacillus plantarum has been observed to play a crucial role in shaping the sensory properties of chili sauce. However, the specific taste-active metabolites responsible for the desirable flavor profile of chili sauce remain inadequately characterized. This study employed a combination of metabolomics and web-based computational tools analysis to investigate the dynamic changes in taste-active metabolites during chili sauce fermentation. Initially, metabolites were rapidly annotated using a feature-based molecular network, leading to the tentative annotation of 206 metabolites, of which a significant proportion had not been previously reported. Subsequently, the VirtualTaste tool identified dihydrosphingosine, lactic acid, isoleucine, phytosphingosine, and gluconic acid as potential taste markers for quality control. Finally, pathway enrichment analysis revealed that these components were primarily associated with amino acid tRNA, phenylalanine, tyrosine, and tryptophan biosynthesis, as well as sphingolipid metabolism. This study provides valuable insights into the mechanisms underlying the formation of the distinctive flavor of chili sauce.

Deep learning (DL) models have been recently widely used to extract task-oriented patterns from large scale of datasets, and to improve the data image understanding and analysis accuracy in many different decision-making processes for tasks such as image classification, segmentation, detection, and so on. However, in practice, the performances of DL models are easily affected by environmental illumination conditions. Conversely, DL models can also be utilized for extracting the illumination hints from the images, and these hints are critically useful for improving the model robustness, classifying the environmental scenes, estimating scene depth information, and rendering 3D objects. In this study, an extensive and exhaustive review is carried out for DL based color constancy, indoor and outdoor illumination estimation, and image depth estimations with the considerations of strengths and weaknesses of DL models. This study also explores the different network designs and the paradoxes in parameter optimization during the model training. Current technology barriers involved in implementing these models and recommendations to overcome these barriers are also suggested in the review.

Kubernetes (K8s) defines standardized APIs for container-based cluster orchestration so it becomes possible for application managers to deploy their applications in a unified manner across different cloud providers. A practical problem is however feature incompatibility between different K8s vendors, who offer commercial K8s products based on the open-source K8s distribution. A large number of documented features in this open-source distribution are optional features that are turned off by default, but can be activated by setting specific combinations of parameters and plug-in components in configuration manifests for the K8s control plane and worker node agents. However, none of these configuration manifests are standardized, giving K8s vendors the freedom to hide the manifests behind a single, more restricted, and proprietary customization interface. Therefore some optional K8s features cannot be activated consistently across K8s vendors and applications that require these features cannot be run on those vendors. In this paper we present a unified, vendor-agnostic feature management approach that bypasses the proprietary customization interface of K8s vendors in order to consistently activate optional K8s features across a federation of clusters hosted by different Kubernetes vendors. We describe vendor-agnostic reconfiguration tactics that are already applied in industry and cover a wide range of optional K8s features. Based on these tactics, we design and implement an autonomic controller for declarative feature compatibility management across a cluster federation. We found that the features configured through our vendor-agnostic approach have no impact on application performance when compared with a cluster where the features are configured using the configuration manifests of the open-source K8s distribution. Moreover, the maximum time to complete reconfiguration of a single feature is within 100 seconds, which is 6 times faster than using proprietary customization interfaces of mainstream K8s vendors such as Google Kubernetes Engine. However, there is a non-negligible disruption to running applications when performing the reconfiguration; this disruption impact does not appear using the proprietary customization methods of the K8s vendors. Therefore, our approach is best applied in the following three use cases: (i) when starting up new K8s clusters, (ii) when optional K8s features of existing clusters must be activated as quickly as possibly and temporary disruption to running applications can be tolerated or (iii) when proprietary customization interfaces do not allow to activate the desired optional feature.

Genomics involving tens of thousands of genes is a complex system determining phenotype. An interesting and vital issue is that how to integrate highly sparse genetic genomics data with a mass of minor effects into prediction model for improving prediction power. We find that deep learning method can work well to extract features by transforming highly sparse dichotomous data to lower dimensional continuous data in a non-linear way. This idea may provide benefits in risk prediction based on genome-wide data associated e.g. integrating most of the information in the genotype data. Hence, we developed a multi-stage strategy to extract information from highly sparse binary genotype data and applied it for risk prediction. Specifically, we first reduced the number of biomarkers via a univariable regression model to a moderate size. Then a trainable auto-encoder was used to extract compact representations from the reduced data. Next, we performed a LASSO problem process over a grid of tuning parameter values to select the optimal combination of extracted features. Finally, we applied such feature combination to two prognostic models, and evaluated predictive effect of the models. The results of simulation studies and real data applying indicated that these highly compressed transformation features could better improve predictive performance and did not easily lead to over-fitting.

This paper describes an image processing method that makes use of image parts instead of neural parts. Neural networks excel at image or pattern recognition and they do this by constructing complex networks of weighted values that can cover the complexity of the pattern data. These features however are integrated holistically into the network, which means that they can be difficult to use in an individual sense. A different method might scan individual images and use a more local method to try to recognise the features in it. This paper suggests such a method, where a trick during the scan process can not only recognise separate image parts, as features, but it can also produce an overlap between the parts. It is therefore able to produce image parts with real meaning and also place them into a positional context. Tests show that it can be quite accurate, on some handwritten digit datasets, but not as accurate as a neural network, for example. The fact that it offers an explainable interface could make it interesting however. It also fits well with an earlier cognitive model, and an ensemble-hierarchy structure in particular.

Epigenetic aging has been found associated with a number of phenotypes and diseases. Few studies investigated its effect on lung function in relatively older people. However, this effect has not been explored in younger population. This study examines whether lung function at adolescent can be predicted with epigenetic age accelerations (AAs) using machine learning techniques. DNA methylation based AAs were estimated in 326 matched samples at two time points (at 10 years and 18 years) from the Isle of Wight Birth Cohort. Five machine learning regression models (linear, lasso, ridge, elastic net, and Bayesian ridge) were used to predict FEV1 (Forced Expiratory Volume in one second) and FVC (Forced Vital Capacity) at 18 years from feature selected predictor variables (based on mutual information) and AA changes between the two time points. The best models were ridge regression (R2 = 75.21% ± 7.42%; RMSE = 0.3768 ± 0.0653) and elastic net regression (R2 = 75.38% ± 6.98%; RMSE = 0.445 ± 0.069) for FEV1 and FVC, respectively. This study suggests that the application of machine learning in conjunction with tracking changes in AA over life span can be beneficial to assess the lung health in adolescence.

MicroRNAs are used as biomarkers for classification of cancer subtypes since certain miRNAs are differentially expressed in normal and patient samples. Moreover, miRNAs target mRNAs and can heavily influence Gene Expressions. Thus, deregulation of miRNAs is linked to various disorders. Thus, miRNAs can be used for prognosis and developing personalized health solutions for patients. Given the importance of miRNAs, there has been substantial work done in the field. In this paper, recent works in the field of using miRNAs expressions of patients were considered. A total of 20 papers were surveyed which utilized feature selection ensembles, fuzzy logic as well as deep learning. 10 papers have been reported which offer insight into how miRNAs can be utilized for subtype-specific or generalized cancer diagnosis.

Despite the long and extensive history of 3D technology, it has recently attracted the attention of researchers. This technology has become the center of interest of young people because of the real feelings and sensations it creates. People see their environment as 3D because of their eye structure. In this study, it is hypothesized that people lose their perception of depth during sleepy moments and that there is a sudden transition from 3D vision to 2D vision. Regarding these transitions, the EEG signal analysis method was used for deep and comprehensive analysis of 2D and 3D brain signals. In this study, a single-stream anaglyph video of random 2D and 3D segments was prepared. After watching this single video, the obtained EEG recordings were considered for two different analyses: the part involving the critical transition (transition-state) and the state analysis of only the 2D versus 3D or 3D versus 2D parts (steady-state). The main objective of this study is to see the behavioral changes of brain signals in 2D and 3D transitions. To clarify the impacts of the human brain’s power spectral density (PSD) in 2D-to-3D (2D_3D) and 3D-to-2D (3D_2D) transitions of anaglyph video, 9 visual healthy individuals were prepared for testing in this pioneering study. Spectrogram graphs based on Short Time Fourier transform (STFT) were considered to evaluate the power spectrum analysis in each EEG channel of transition or steady-state. Thus, in 2D and 3D transition scenarios, important channels representing EEG frequency bands and brain lobes will be identified. To classify the 2D and 3D transitions, the dominant bands and time intervals representing the maximum difference of PSD were selected. Afterward, effective features were selected by applying statistical methods such as standard deviation (SD), maximum (max), and Hjorth parameters to epochs indicating transition intervals. Ultimately, k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA) algorithms were applied to classify 2D_3D and 3D_2D transitions. The frontal, temporal, and partially parietal lobes show 2D_3D and 3D_2D transitions with a good classification success rate. Overall, it was found that Hjorth parameters and LDA algorithms have 71.11% and 77.78% classification success rates for transition and steady-state, respectively.

The aim of this study was to examine brightness effect, which is the perceptual property of visual stimuli, on brain responses obtained during visual processing of these stimuli. For this purpose, brain responses of the brain to changes in brightness were explored comparatively using different emotional images (pleasant, unpleasant and neutral) with different luminance levels. Moreover, electroencephalography recordings from 12 different electrode sites of 31 healthy participants were used. The power spectra obtained from the analysis of the recordings using short time Fourier transform were analyzed, and a statistical analysis was performed on features extracted from these power spectra. Statistical findings obtained from electrophysiological data were compared with those obtained from behavioral data. The results showed that the brightness of visual stimuli affected the power of brain responses depending on frequency, time and location. According to the statistically verified findings, the distinctive effect of brightness occurred in the parietal and occipital regions for all the three types of stimuli. Accordingly, the increase in the brightness of pleasant and neutral images increased the average power of responses in the parietal and occipital regions whereas the increase in the brightness of unpleasant images decreased the average power of responses in these regions. However, the increase in brightness for all the three types of stimuli reduced the average power of frontal and central region responses (except for 100-300 ms time window for unpleasant stimuli). The statistical results obtained for unpleasant images were found to be in accordance with the behavioral data. The results also revealed that the brightness of visual stimuli could be represented by changing the activity power of the brain cortex. The main contribution of this research was to comprehensively examine brightness effect on brain activity for images with different emotional content and different frequency bands at different time windows of visual processing for different brain regions. The findings emphasized that the brightness of visual stimuli should be viewed as an important parameter in studies using emotional image techniques such as image classification, emotion evaluation and neuro-marketing.

Spatial data conflation plays a fundamental role in in many aspects of modern Geographic Information Systems (GIS) research and development such as geospatial data visualization, incremental updating of databases and disaster evaluation. The primary objective of conflation is to derive valuable information based on the comparison of multiple spatial data sources of homogeneous or heterogeneous nature. This paper reviews the state of art of the concept of conflation, feature matching, the most important progresses of conflation between image and road network data, and the complexity of spatial conflation.

This paper proposes a crack recognition method based on high-resolution line array cameras and adaptive lifting algorithm. By defining the crack rate, this algorithm calculates the ratio of the crack area to the area of the entire collected image to characterize the damage extent of the current section. The algorithm first uses image preprocessing to reduce the image noise, then uses histogram equalization to enhance the feature of the crack region, divides the whole image into multiple sub-blocks, and extracts region features in the sub-block. At the same time, this algorithm defines related feature descriptors, and constructs weak classifiers according to each feature descriptor, and converts the weak classifiers into strong classifiers by using an adaptive lifting algorithm. Finally, this algorithm realizes the division of the crack regions. Experimental results show that the proposed algorithm can meet the actual needs and is better than other classical algorithms.

In recent years, many algorithms based on end-to-end deep networks have been proposed to deal with the target detection problem of videos. However, the deep network models usually consume a lot of computing resources during the procedure of analysis of videos with complex dynamic backgrounds. In this paper, a new method of object detection based on information theory is presented. Firstly, each frame in a video is converted into an effective information map by using the Harris corner detection method. Secondly, the sensitive areas in the frame are extracted by using the context information and the effective information maps of the consecutive video frames. The sensitive areas in the video frame are the candidate areas where the target objects would be appeared at high probabilities. Thirdly, the information entropy features of each sensitive area are extracted to form the feature matrix, based on which, an SVM model is trained for selecting the target areas from the sensitive areas. Finally, the locations of the objects are detected based on the target areas in the video with a complex dynamic background. As a lightweight video detection framework, the method presented in this paper can save a lot of computing resources. Experimental results show that this method can achieve good results in the benchmark of CDnet 2014.

The information of seismic damage of buildings in SAR images of different time phase, especially in SAR images after earthquake, is easily disturbed by other factors, which affects the accuracy of information discrimination. In order to identify and evaluate the distribution information of the seismic damage accurately and make full use of the abundant texture features in the SAR image. The conventional method of change detection based on texture features usually takes the pixel as the calculating unit. In this paper, a method of texture feature change detection of SAR images based on watershed segmentation algorithm is proposed. Based on the optimization of texture feature parameters, the feature parameters are segmented by the watershed segmentation algorithm, and the feature object image is obtained. This method introduces the idea of object oriented, and carries out the calculation of the difference map at the object level, Finally, the classification threshold value of different types of seismic damage types is selected, and the recognition of building damage is achieved. Taking the ALOS data before and after the earthquake in Yushu as an example to verify the effectiveness of the method, the overall accuracy of the building extraction is 88.9%, Compared with pixel-based methods, it is proved that the proposed method is effective.

As people grow a custom to effortless outdoor navigation there is a rising demand for similar possibility indoors as well. Unfortunately, indoor localization, being one of the necessary requirements for navigation, continues to be problem without a clear solution. In this article we are proposing a method for an indoor positioning system using a single image. This is made possible using small preprocessed database of images with known control points as the only preprocessing needed. Using feature detection with SIFT algorithm we can look through the database and find image which is the most similar to the image taken by user. Pair of images is then used to find coordinates of database image using PnP problem. Furthermore, projection and essential matrices are determined allowing for the user image localization ~ determining the position of the user in indoor environment. Benefits of this approach lies in the single image being the only input from user and no requirements for new onsite infrastructure and thus enables a simpler realization for the building management.

Speech emotion recognition (SER) is important in facilitating natural human-computer interactions. In speech sequence modelling, a vital challenge is to learn context-aware sentence expression and temporal dynamics of para-linguistic features to achieve unambiguous emotional semantic understand-ing. In previous studies, the SER method based on the single-scale cascade feature extraction module could not effectively preserve the temporal structure of speech signals in the deep layer, downgrading the sequence modeling performance. In this paper, we propose a novel multi-scale feature pyramid network to mitigate the above limitations. With the aid of the bi-directional feature fusion of the pyramid network, the emotional representation with adequate temporal semantics is obtained. Experiments on the IEMOCAP corpus demonstrate the effectiveness of the proposed methods and achieve competitive results under speaker-independent validation.

Vehicle logo detection plays a crucial role in various computer vision applications, such as vehicle classification and detection. In this research, we propose an improved vehicle logo detection method leveraging the MAMBA structure. The MAMBA structure integrates multiple attention mechanisms and bidirectional feature aggregation to enhance the discriminative power of the detection model. Specifically, we introduce the Multi-Head Attention for Multi-Scale Feature Fusion (MHAMFF) module to capture multi-scale contextual information effectively. Moreover, we incorporate the Bidirectional Aggregation Mechanism (BAM) to facilitate information exchange between different layers of the detection network. Experimental results on a benchmark dataset (VLD-45 dataset) demonstrate that our proposed method outperforms baseline models in terms of both detection accuracy and efficiency. Furthermore, extensive ablation studies validate the effectiveness of each component in the MAMBA structure. Overall, the proposed MAMBA-based vehicle logo detection approach shows promising potential for real-world applications in intelligent transportation systems.

In response to the problem of poor detection ability of object detection models for small-scale targets in intelligent transportation scenarios, a fusion method is proposed to enhance the features of small-scale targets, starting from feature utilization and fusion methods. The algorithm is based on the YOLOv4 tiny framework and enhances the utilization of shallow and mid level features on the basis of FPN, improving the detection accuracy of small and medium-sized targets; In view of the problem that the background of the intelligent traffic scene image is cluttered and there is more redundant information, the CBAM attention module is used to improve the attention of the model to the traffic target; To address the problem of data imbalance and prior bounding box adaptation in custom traffic datasets that expand traffic images in COCO and VOC, we propose a Copy Paste method with improved generation method and a K-means algorithm with improved distance measurement to enhance the model's detection ability for corresponding categories. Comparative experiments were conducted on a customized 260 thousand traffic dataset containing public traffic images, and the results showed that compared to YOLOv4 tiny, the proposed algorithm improved mAP by 4.9% while still ensuring the real-time performance of the model.

Link prediction is a significant field in network science, which focuses on predicting the probability of the existence or formation of a link between nodes in a social network based on currently observed connections. The effectiveness of traditional algorithms can vary depending on the type of network, making some methods more suitable than others for specific scenarios. Recently, several efficient link prediction algorithms have been developed, demonstrating robust results in both prediction accuracy and interpretability. However, existing research has not clearly established the relationship between network characteristics and link creation mechanisms for community influence analysis and anomaly detection. The ability to predict complex networks with diverse features still requires further investigation. In light of this, we introduce a novel framework designed to combine the best features of different link prediction algorithms when applied to the network, with the aim of achieving more reliable predictions about how networks will evolve. According to the proposed framework, we first focus on the feature extraction stage. During this phase, we systematically identify and extract a comprehensive set of features from the network before moving onto the classification phase. Here, we utilize state-of-the-art ensemble learning models to assess and classify potential links within the network. By training our machine learning models on the extracted features, we can effectively predict whether a particular link is likely to form (positive link) or unlikely to form (negative link). The ML models were trained and evaluated using two datasets: Twitch and Facebook. Results: Additionally, we assessed their performance on these datasets by conducting specific preprocessing and hyperparameter tuning steps. The final prediction model achieves AUC values between 94% and 99.3%. This research contributes significantly to efforts aimed at enhancing link prediction in dynamic socisal network contexts, providing valuable insights into the effectiveness of different ML algorithms in predicting future connections and enhancing our understanding of network dynamics.

Marine oil spills seriously threaten the coastal environment, challenge the safety of marine eco-system, and cause huge economic losses to marine fishery, aquaculture, and tourism industries. Timely and accurate oil spill monitoring and early warning can not only make the oil spill acci-dent be handled timely and effectively but also provide key information for investigating the re-sponsibility of the polluters. Polarimetric synthetic aperture radar (SAR) has shown its great po-tential in oil spill classification for its capacity to distinguish mineral oil and biogenic look-alikes. Traditional oil spill classification approaches rely on the scattering matrix elements or pre-designed scattering mechanism-based polarimetric SAR features, which limits its robustness and generalization in large-scale marine surveillance tasks. The recent development of deep learning frameworks, especially the convolutional neural network (CNN) provides a powerful tool for remote sensing image classification. However, conventional CNN only accepts re-al-valued input, which cannot efficiently analyze the polarimetric information carried by com-plex-valued SAR data. In this study, a new framework that employs the complex-valued neural network was proposed for marine oil spill detection. Sophisticated studies were designed to evaluate the contribution of the physical scattering mechanism-based features, complex-valued framework, and convolutional feature extraction structures respectively. Comparative studies were conducted on four study cases using Radarsat-2 and SIR-C fully polarimetric SAR data. The outcome of the proposed study may shed light on the development of an end-to-end marine spill classification system and enlarge its application in practical marine surveillance operations.

Convolutional neural networks (CNNs) face challenges in capturing long-distance text correlations, and Bidirectional Long Short-Term Memory (BiLSTM) networks exhibit limited feature extraction capabilities for text classification of work order. To address the abovementioned problems, this work utilizes an ensemble learning approach to integrate model elements efficiently. This study presents a method for classifying work order texts using a hybrid neural network model called BERT-BiLSTM-CNN. First, use Bert for preprocessing to obtain text vector representations. Then, capture context and process sequence information through BiLSTM. Next, capture local features in the text through CNN. Finally, obtain classification results through Softmax. Through comparative analysis, the method of fusing these three models is superior to other hybrid neural network model architectures in multiple classification tasks. It has a significant effect on work order classification.

Partial Differential Equations (PDEs) constitute a fundamental framework for modeling various physical phenomena across diverse fields such as physics, engineering, and finance. Solving PDEs accurately and efficiently remains a significant challenge, particularly when dealing with complex systems or high-dimensional data. In this paper, we present a novel approach based on deep learning techniques to investigate the effective factors in solving PDEs. Our methodology leverages the expressive power of deep neural networks to learn and represent the underlying relationships between input parameters and the solutions of PDEs. By incorporating domain knowledge into the network architecture and training process, our approach aims to provide insights into the key factors influencing the solution of PDEs. We demonstrate the effectiveness of our method through numerical experiments on various PDEs, showcasing its capability to identify important features and improve predictive accuracy. Our findings suggest that deep learning techniques offer a promising avenue for understanding and analyzing the complex dynamics inherent in PDEs, paving the way for enhanced computational methods in scientific and engineering applications.

Due to the significant price difference, expensive oils like olive oil are often blended with cheaper edible oils. This practice of adulteration in edible oils, aimed at increasing profits for producers, poses a major concern for consumers. Furthermore, adulteration in edible oils can lead to various health issues impacting consumer well-being. In order to meet the requirements of fast, non-destructive, universal, accurate and reliable quality testing for edible oil, the oblique-incidence reflectivity difference (OIRD) method combined with machine learning algorithms was introduced to detect a variety of edible oils. The prediction accuracy of Gradient Boosting, K-Nearest Neighbor, and Random Forest models exceeded 95%. Experimental results indicate that the OIRD method can serve as a powerful tool for detecting edible oils.

Pedestrian Dead Reckoning (PDR) is a promising algorithm for indoor positioning. However, the accuracy of PDR degrades due to the accumulated error, especially in multi-floor buildings. This paper introduces a three-dimensional (3D) positioning method based on terrain feature matching to reduce the influence of accumulated error in multi-floor scene. The proposed method involves two steps: motion pattern recognition and position matching calibration. The motion pattern recognition aims to detect different motion states, i.e., taking the stairs or horizontal walking, from the streaming data. Then, stair entrances and corridor corners are matched with transition points of motion states and pedestrian turning points, respectively. After matching, calibration is performed to eliminate the accumulated errors. By carrying out experiments on a two-floor closed-loop path with a walking distance about 145 m, it is shown that this method can effectively reduce the accumulated error of PDR , achieving accurate 3D positioning . The average error is reduced from 6.60 m to 1.83 m.

We propose Context-aware Feature Transformer Network (CaFTNet), a novel network for human pose estimation. To address the issue of limited modeling of global dependencies in convolutional neural networks, we design Transformerneck to strengthen the expressive power of features. Transformerneck directly substitutes the 3×3 convolution in bottleneck of HRNet with Contextual Transformer (CoT) block, while reducing the complexity of the network. Specifically, CoT first produces keys with static contextual information through 3×3 convolution. Then, relying on the query and contextualization keys, the dynamic contexts are generated through two concatenated 1×1 convolutions. Static and dynamic contexts are eventually fused as an output. Additionally, for the multi-scale networks, in order to further refine the features of the fusion output, we propose an Attention Feature Aggregation Module(AFAM). Technically, given an intermediate input, AFAM successively deduces attention maps along channel and spatial dimensions. Then, Adaptive refinement module(ARM) is exploited to activate the obtained attention maps. Finally, the input undergoes adaptive feature refinement through multiplication with the activated attention maps. Through the above studies, our lightweight network provides a powerful clue for detection of keypoints. Experiments are implemented on the COCO and MPII datasets. The model achieves 76.2 AP on the COCO val2017. Compared to other methods with the CNN as the backbone, CaFTNet reduces the number of parameters by 72.9 %. On the MPII, our method uses only 60.7% of the number of parameters, acquiring semblable results to other methods with the CNN as the backbone.

Broadly speaking, binominal and biverbal lexical constructions have been studied independently in different research traditions and frameworks. It is true that the two do not necessarily have overlapping areal distributions, but the fundamental question remains whether Indo-European NN compounds and Transeurasian VV compounds have nothing in common. From a purely theoretical perspective also, the cross-categorial parallelism is worthy of serious investigation because a theory of the process of compounding that inherits the spirit of X-bar theory strives to capture commonalities among various compounding patterns and category-specific constraints observed within and across languages. Against this background, a cross-categorial comparison, not within but across languages, is made of coordinative binominal and biverbal constructions. NN and VV coordinate compounds from English and Japanese are examined in detail using the methodology of contrastive morphology and decompositional lexical semantics. It is shown that dvandva is possible not only in NN but also in VV coordinate compounds and furthermore that the dvandva-appositive distinction in NN coordinate compounds recurs in VV coordinate compounds.

Detecting rotational objects in remote sensing imagery is a significant challenge. These images typically encompass a broad field of view, featuring diverse and intricate backgrounds, with ground objects of various sizes densely scattered. As a result, identifying objects of interest within these images is a daunting task. While the integration of Convolutional Neural Networks (CNN) and Transformer networks leads to some advancements in rotational object detection, there is still room for improvement, particularly in enhancing the extraction and utilization of information related to smaller objects. To address this, our paper presents a multi-scale feature fusion module and a global feature context aggregation module. Initially, we fuse original, shallow, and deep features to reduce the loss of shallow feature information, thereby improving the detection performance of small objects in complex backgrounds. Subsequently, we compute the correlation of contextual information within feature maps to extract valuable insights. We name the newly proposed model the "Multiscale Feature Context Aggregation Module" (MFCA). We evaluate our proposed methodology on three challenging remote sensing datasets: DIOR-R, HRSC, and MAR20. Comprehensive experimental results show that our approach surpasses baseline models by 2.07\% mAP, 1.02\% mAP, and 1.98\% mAP on the DIOR-R, HRSC2016, and MAR20 datasets, respectively.

Agriculture is pivotal in national economies, with pest detection significantly influencing food quality and quantity. Pest classification remains challenging in automated agriculture monitoring systems, exacerbated by the non-uniform pest scales and the scarcity of high-quality datasets. In this study, we constructed a pest dataset by acquiring domain-agnostic images from the Internet and resizing them to a standardized 299x299 pixel format. Additionally, we employed diffusion models to generate supplementary data. While Convolutional Neural Networks (CNNs) are prevalent for prediction and classification, they often lack effective global information integration and discriminative feature representation. To address these limitations, we propose the RS Transformer, an innovative model that combines elements like the Region Proposal Network, Swin Transformer, and ROI Align. Additionally, we introduce the Randomly Generated Stable Diffusion Dataset (RGSDD) to augment the availability of high-quality pest datasets. Extensive experimental evaluations demonstrate the superiority of our approach compared to both two-stage models (SSD and Faster R-CNN) and one-stage models (YOLOv3, YOLOv4, YOLOv5m, YOLOv8, and DETR). We rigorously assess performance using metrics such as mean Average Precision (mAP), F1Score, Recall, and mean Detection Time (mDT). Our research contributes to advancing pest detection methodologies in automated agriculture systems, promising improved food production and quality.

The extraction of roof feature lines is an important foundation for realizing large-scale and batch 3D modeling. However, the current traditional point cloud segmentation algorithms do not have satisfactory results in extracting roof feature lines of Chinese traditional residential buildings. In this paper, taking Jingping Village in Western Hunan as an example, we propose a method that combines multiple algorithms based on slope segmentation of roof patches to extract feature lines. Firstly, VDVI and CSF algorithms are used to extract the building and roof point cloud based on the MVS point cloud. Secondly, according to roof features, village buildings are classified, and 3D roof point cloud is projected into 2D regular grid data. Finally, the roof slope is segmented via slope direction, and internal and external feature lines are obtained after refinement through Canny edge detection and Hough straight line detection. Results reveal that this method effectively extracts feature lines of low-building roofs in traditional villages, with slope-based roof surface segmentation accuracy surpassing 99.6%. This method significantly outperforms the RANSAC algorithm and region segmentation algorithm.

Recently, deep learning models have been widely applied to modulation recognition, which have become a hot topic due to their excellent end-to-end learning capabilities. However, current methods are mostly based on uni-modal inputs, which suffer from incomplete information and local optimization. To complement the advantages of different modalities, we focus on the multi-modal fusion method. Therefore, we introduce an iterative dual-scale attentional fusion (iDAF) method to integrate multimodal data. Firstly, two feature maps with different receptive field sizes are constructed using local and global embedding layers. Secondly, the feature inputs are iterated into the Iterative Dual Scale Attention Module (iDCAM), where the two branches capture the details of high-level features and the global weights of each modal channel, respectively. The iDAF not only extracts the recognition characteristics of each specific domains, but also complements the strengths of different modalities to obtain a fruitful view. Our iDAF achieves a recognition accuracy of 93.5\% at 10dB and 0.6232 at full SNR. The comparative experiments and ablation studies effectively demonstrate the effectiveness and superiority of the iDAF.

Cancer metastasis accounts for approximately 90% of cancer deaths, and elucidating markers in metastasis is the first step in its prevention. To characterize metastasis marker genes of breast cancer (MGs), XGBoost models that classify metastasis status were trained with gene expression profiles from TCGA. Then, a metastasis score (MS) was assigned to each gene by calculating the inner product between the feature importance and AUC performance of the models. As a result, the 54, 202, and 357 genes with the highest MS were characterized as MGs by empirical P-value cutoffs of 0.001, 0.005, and 0.01, respectively. The three sets of MGs were compared with those from existing metastasis marker databases, which provided significant results in most comparisons. We noticed that the set of MGs with the median EP cutoff showed better performance than the other two sets, suggesting the importance of the cutoff used in determining MGs. They were also significantly enriched in biological processes associated to breast cancer metastasis. The MGs that could not be identified by statistical analysis (e.g., GOLM1, ELAVL1, UBP1, and AZGP1) as well as the MGs with the highest MS (e.g., ZNF676, FAM163B, LDOC2, IRF1, and STK40) were verified via the literature. Additionally, we checked how close the MGs are located to each other in the protein–protein interaction networks. We expect that the characterized markers will help understand and prevent breast cancer metastasis.

To improve the accuracy of short-term multi-energy load prediction models for integrated energy systems, the historical development law of the multi-energy loads must be considered. Moreover, understanding the complex coupling correlation of the different loads in the multi-energy systems and accounting for other load-influencing factors, such as weather, may further improve the forecasting performance of such models. In this study, a two-stage fuzzy optimization method is proposed for the feature selection and identification of the multi-energy loads. To enrich the information content of the prediction input feature, we introduced a copula correlation feature analysis in the proposed framework, which extracts the complex dynamic coupling correlation of multi-energy loads and applies Akaike information criterion (AIC) to evaluate the adaptability of the different copula models presented. Furthermore, we combined a NARX neural network with Bayesian optimization and an extreme learning machine model optimized using a genetic algorithm to effectively improve the feature fusion performances of the proposed multi-energy load prediction model. The effectiveness of the proposed short-term prediction model was confirmed by the experimental results obtained using the multi-energy load time-series data of an actual integrated energy system.

Unmanned Aerial Vehicles (UAVs) image object detection has great application value in military and civilian fields. However, the objects in the captured images from UAVs have problems of large scale variation, complex backgrounds, and a large proportion of small objects. To resolve these problems, a multi-scale object detector based on coordinate and global information aggregation is proposed, named CGMDet. Firstly, a Coordinate and Global Information Aggregation Module (CGAM) is designed by aggregating local, coordinate, and global information, which can obtain features with richer context information. Secondly, a Multi-Feature Fusion Pyramid Network (MF-FPN) is proposed, which can better fuse features of different scales and obtain features containing more context information through repeated use of feature maps, to better detect multi-scale targets. Moreover, more location information of low-level feature maps is integrated to improve the detection results of small targets. Furthermore, we modified the bounding box regression loss of the model to make the model more accurately regress the bounding box and faster convergence. Finally, the proposed CGMDet was tested on VisDrone and UAVDT datasets and mAP0.5 of 50.9% and 48% was obtained, respectively. At the same time, our detector achieved the best results compared to other detectors.

In this study, we demonstrate how the incorporation of appropriate feature engineering together with the selection of a Machine Learning (ML) algorithm that best suits the available dataset, leads to the development of a predictive model for transformation temperatures that can be applied to a wide range of shape memory alloys. We develop a gradient boosting ML surrogate model capable of predicting Martensite Start, Martensite Finish, Austenite Start, and Austenite Finish transformation temperatures with an average accuracy of more than 95% by explicitly taking care of potential distribution changes when modeling different alloy systems. We included heat treatment, rolling, extrusion processing parameters, and alloy system categorical features in the model input features to achieve more accurate and realistic results. In addition, using Shapley values, which are calculated based on the average marginal contribution of features to all possible coalitions, this study was able to gain insights into the governing features and their effect on predicted transformation temperatures, providing a unique opportunity to examine the critical parameters and features in martensite transformation temperatures.

In recent years, with the development of social media, people are more and more inclined to upload text, pictures and videos on the platform to express their personal emotions, thus the number of short videos is increasing and becoming the first choice for people to socialize. Unlike the traditional way, people can convey their personal emotions and opinions through media other than words, such as video images, etc. for external information. Therefore, the expression and analysis of emotions is not only through text, but also through the analysis of emotional needs in images and videos, and the research scholars have customized products for individual users. Compared with pure text content, video information can more intuitively express users' happiness, anger and sorrow, thus short video-related applications have gained more and more popularity among Internet users in recent years. However, not all short videos on social networking sites can accurately express users' emotions, and related text information can more accurately assist sentiment analysis and thus improve accuracy. However, short video sentiment analysis based on video frame images is inaccurate in some scenarios, such as when expressing tears of joy, the sentiment expressed by the user's facial expression and voice are different, which will cause errors in the analysis of sentiment. As a result, researchers began to consider multimodal sentiment analysis to reduce the impact of the above scenarios on short video sentiment analysis. This paper focuses on proposing a sentiment analysis method for short videos. We first propose a residual attention model to make full use of the information in audio to classify the emotions contained in them. Then the text information in the dataset is classified by feature extraction. The key to extract features from text information is not only to retain the semantic information of the text, but also to explore the potential emotional information in the text, so as to ensure the integrity of the text information features. The experiments show that the sentiment analysis model proposed in this paper is more superior than the baselines.

Despite a rapid rise in NBS development in recent years, the methods for evaluating NBS still have certain gaps. We propose an approach based on a combination of remote sensing data and meteorological variables to reconstruct the spatio-temporal variation of net ecosystem exchange from eddy-covariance stations. Lagrangian particle dispersion model was used for upscaling of satellite images and flux towers. We trained data-driven models based on kernel methods separately for each selected land cover class. The results suggest that the proposed approach to quantifying carbon exchange on a medium-to-large scale by blending eddy covariance flux data with moderate resolution satellite and weather data provides a set of key advantages over previously deployed methods: (1) scalability, achieved via the validation design based on a separate set of eddy covariance stations; (2) high spatial and temporal resolution due to use of the Landsat imagery; (3) robust and accurate predictions due to improved data quality control, advanced machine learning techniques, and rigorous validation. The machine learning models yielded high cross-validation results. Overall we present here globally scaled technology for the land sector based on high resolution remote sensing imagery, meteorological variables, and direct carbon flux measurements of eddy covariance flux stations.

Background: Digital clinical measures collected via various digital sensing technologies such as smartphones, smartwatches, wearables, ingestible, and implantable are increasingly used by individuals and clinicians to capture health outcomes or behavioral and physiological characteristics of individuals. Time series classification (TSC) is very commonly used in modeling digital clinical measures. While deep learning models for TSC are very common and powerful, there exist some fundamental challenges. This review presents non-deep learning models commonly used for time series classification in biomedical applications that achieve high performance. Objective: We performed a systematic review to characterize the techniques used in time series classification of digital clinical measures throughout all stages of data processing and model building. Methods: We conducted a literature search on PubMed, and the Institute of Electrical and Electronics Engineers (IEEE), Web of Science, and SCOPUS databases using a range of search terms to retrieve peer-reviewed articles reporting academic research on digital clinical measures in the five year period between June 2016 and June 2021. We identified and categorized research studies based on the types of classification algorithms and sensor input types. Results: We found 452 papers in total from four different databases: PubMed, IEEE, Web of Science Database, and SCOPUS. After removing duplicates and irrelevant papers, 135 articles remained for detailed review and data extraction. Among these, engineered features using time series methods that were subsequently fed into widely-used machine learning classifiers was the most commonly used technique and also most frequently achieved the best performance metrics (77 out of 135 articles). Statistical modeling (24 out of 135 articles) algorithms were the second most common and also second best classification technique. Wavelet-based classification models (8 out of 135 articles) were also common. Electroencephalogram (29 out of 135 articles) was the most common data type used as an input. Accuracy was the most commonly reported performance metric, with 67.65% of articles reporting on accuracy. In this review paper, we provide summaries of signal pre-processing methods, feature engineering and selection methods, time series models, as well as model interpretations. Importantly, we found that about 50% of the papers only report one performance metric, which may result in a skewed view of overall performance. Conclusion: While high time series classification performance has been achieved in digital clinical, physiological, or biomedical measures, no standard benchmark datasets, modeling methods, or reporting methodology exist. There is no single widely used method for time series model development or feature interpretation– many different methods have proven successful.

COVID-19 vaccine hesitancy is considered responsible for the lower rate of acceptance of vaccines in many parts of the world. However, sources of this hesitancy are rooted in many social, political, and economic factors. This paper strives to find the most important variables in predicting the COVID-19 vaccination uptake. We introduce an explainable machine learning (ML) framework to understand the COVID-19 vaccination uptake around the world. To predict vaccination uptake, we have trained a random forest (RF) regression model using a number of sociodemographic and socioeconomic data. The traditional decision tree (DT) regression model is also implemented as the baseline model. We found that the RF model performed better than the DT model since RF is more robust to handle nonlinearity and multi-collinearity. Also, we have presented feature importance based on impurity measure, permutation, and Shapley values to provide the most significant unbiased features. It is found that electrification coverage and Gross Domestic Product are the strongest predictors for higher vaccination uptake, whereas the Fragile state index (FI) contributed to lower vaccination uptake. These findings suggest addressing issues that are found responsible for lower vaccination uptake to combat any future public health crisis.

Karst wetlands are being seriously damaged, and protecting it has become an important matter. Karst vegetation is the essential component of wetland and plays an important role in in the ecological functions of wetland ecosystems. Classifying karst vegetation is important for karst wetlands protection and management. This paper addressed to classify karst vegetation in Huixian National Wetland Park, located in China using the improved SegNet Deep-Learning Algorithm and UAV images. This study proposed a method to fuse single-class SegNet models using the maximum probability algorithm for karst vegetation classification, and compared with object-based RF classification and multi-class SegNet classification, respectively. This paper evaluated the performance of multi-class SegNet model and fusion of single-class SegNet model with different EPOCH values for mapping karst vegetation. A new optimized post-classification algorithm was proposed to eliminate the stitching traces caused by SegNet model prediction. The specific conclusions of this paper include the followings:(1) fusion of four single-class SegNet models produced better classification for karst wetland vegetation than multi-class SegNet model, and achieved the highest overall classification accuracy (87.34%); (2) The optimized post-classification algorithm was able to improve prediction accuracy of SegNet model, and it could eliminate splicing traces; (3) The karst wetland vegetation classifications produced by single-class SegNet model outperformed multi-class SegNet model, and improved classification accuracy(F1-Score) between 10%~25%;(4)The EPCOH values and textural feature important impact on karst wetland vegetation classifications. The SegNet model with EPCOH 15 achieved greater classification accuracy(F1-Score) than the model with EPOCH 5 or 10. The textural feature improved improves the capability of the SegNet model for mapping karst vegetation;(5) Fusion of single-class SegNet models and object-based RF model could provide high classifications results for karst wetland vegetation, and both achieved greater 87% overall accuracy.

Nowadays, speed up development and use of digital devices such as smartphones have put people at risk of internet crimes. The evidence of present crimes in a computer file can be easily unreachable by changing the prefix of a file or other algorithms. In more complex cases, either file divided into different parts or the parts of a file that has information about the file type are deleted, where the file fragment recognition issue is discussed. The known files are divided into different fragments, and different classification algorithms to solve the problems of file fragment recognition. A confusion matrix measures the accuracy of type recognition. In the present study, first, the file is divided into different fragments. Then, the file fragment features, which are obtained from Binary Frequency Distribution (BFD), are reduced by 2 feature reduction algorithms; Sequential Forward Selection algorithm (SFS) as well as Sequential Floating Forward Selection algorithm (SFFS) to delete sparse features that result in increased accuracy and speed. Finally, the reduced features are given to 3 classifier algorithms, Multilayer Perceptron (MLP), Support Vector Machines (SVM), and K-Nearest Neighbor (KNN) for classification and comparison of the results. In this paper, we proposed the algorithm of file type recognition that can recognize 6 types of useful files ( pdf, txt, jpg, doc, html, exe), which may distinguish a type of file fragments with higher accuracy than the similar works done.

Diesel engines, as power equipment, are widely used in the fields of automobile industry, ship and power equipment. Due to wear or faulty adjustment, the valve train clearance abnormal fault is a typical failure of diesel engines, which may result in the performance degradation, even valve fracture and cylinder hit fault. However, the failure mechanism features mainly in time domain and angular domain, on which the current diagnosis methods based, are easily affected by working conditions or hard to extract accurate enough, as the diesel engine keeps running in transient and non-stationary process. This work arms at diagnosing this fault mainly based on frequency band features which would change when the valve clearance fault occurs. For the purpose of extracting a series of frequency band features adaptively，a decomposition technique based on improved variational mode decomposition is investigated in this work. As the connection between the features and the fault is fuzzy, the random forest algorithm is used to analyze the correspondence between features and faults. In addition, the feature dimension is reduced to improve the operation efficiency according to importance score. The experimental results under variable speed condition show that the method based on variational mode decomposition and random forest is capable to detect valve clearance fault effectively.

This paper presents a novel method to extract local features, which instead of calculating local extrema computes global maxima in a discretized scale-space representation. To avoid obtaining precise scales by interpolation and to achieve perfect rotation invariance, two essential techniques, increasing the width of kernels in pixel and utilizing disk-shaped convolution template are adopted in this method. Since the size of a convolution template is finite and finite templates can introduce computational error into convolution, we sufficiently discuss this problem and work out an upper bound of the computational error. The upper bound is utilized in the method to ensure that all features obtained are computed under a given tolerance. Besides, the technique of relative threshold to determine features is adopted to reinforce the robustness for the scene of changing illumination. Simulations show that this new method attains high performance of repeatability in various situations including scale change, rotation, blur, JPEG compression, illumination change and even viewpoint change.

This paper considers target characterization and recognition in radar images with keypoint-based local descriptor. Most of the preceding works rely on the global features or raw intensity values, and hence produce the limited recognition performance. Moreover, the global features are sensitive to the real-world sources of variability, such as aspect view, configu-ration, and incidence angle changes, clutter, articulation, and occlusion. Keypoint-based local descriptor was developed as a powerful strategy to address invariance to contrast change and geometric distortion. This property inspires us to investigate whether the family of local features are relevant for radar target recognition. Most of the preceding works typically devote to finding the correspondences between a collected image and a reference one. The representative applications include image register and change detection. Little work was pursued to target recognition in SAR images. This is because the huge number of local descriptors resulting from radar images make the computational cost and memory consumption unacceptable. To handle the problems, this paper develops two families of methods. The proposed methods are used to achieve target recognition by means of local descriptors. Our first solver refers to building multiple linear regression models, and addresses the problem by the theory of sparse representation. The second scheme rebuilds a new feature by the feature quantization skill, from which the inference can be drawn. Multiple comparative studies are pursued to verify the performance of detectors and descriptors popularly used. The source code was publicly released on https://ganggangdong.github.io/homepage/.

Automated Electrocardiogram (ECG) processing is an important technique which helps in identifying abnormalities in the heart before any formal diagnosis. This research presents a real-time and lightweight R-assisted feature extraction algorithm and a heartbeat classification scheme which achieves highly accurate abnormality detection. In the proposed algorithm, we extract fifteen features from each heartbeat taken from raw Lead-II ECG signals. The features carry medically valuable information such as locations, amplitude and energy of ECG waves (P, Q, R, S, T waves) which are then used for detection of any abnormality that might be present in the heartbeat using various classification algorithms. We have used four popular databases from Physionet and extracted ten thousand ECG signals from each for training the models and benchmarking results. Four classification models i.e. Naïve Bays, k-Nearest Neighbor, Neural Network, Decision Tree were used for abnormality detection validating the efficiency of the system.

The vibration signal of the engine contains strong background noise and many kinds of modulating components, which is difficult to diagnose. Variational mode decomposition (VMD) is a recently introduced adaptive signal decomposition algorithm with a solid theoretical foundation and good noise robustness compared with empirical mode decomposition (EMD). VMD can effectively avoid endpoint effect and modal aliasing. However, VMD cannot effectively eliminate the random noise in the signal, so the random decrement technique is introduced to solve the problem. Based on the crankshaft bearing fault simulation experiment, the four kinds of wear state vibration signals are decomposed by VMD, and the modal components with smaller permutation entropy are selected as fault components. Then the fault component is processed by the random decrement technique, and the Hilbert envelope spectrum of the fault component is obtained. Compared with the fault feature extraction method based on EMD and EEMD, the feature extraction results of the proposed method are better than those of the above two methods. The simulation analysis and the simulation test of the crankshaft bearing fault verify the effectiveness of the proposed method.

Accurate assessment of the breast cancer deterioration degree plays a crucial role in making medical plan, and the important basis for degree assessment is the number of mitoses in a given area of the pathological image. We utilized deep multi-scale fused fully convolutional neural network (MFF-CNN) combing with conditional random felid (CRF) to detect mitoses in hematoxylin and eosin stained histology image. Analyze the characteristics of mitotic detection ----scale invariance and sparsity, as well as the difficulties ---- small amount of data , inconsistent image staining and sample class unbalanced. Based on this, mitotic detection model is designed. In this paper, a tissue-based staining equalization method is used, and to establish an effective training sample set, we select training samples by using CNN. A mitotic detection model fusing multi-level and multi-scale features and context information was designed, and the corresponding training strategy was made to reduce over-fitting. As preliminarily validated on the public 2014 ICPR MITOSIS data, our method achieves a better performance in term of detection accuracy than ever recorded for this dataset.

Human visual system (HVM) can quickly localize the most salient object in scenes, which has been widely applied on natural image segmentation [15]-[19]. In ultrasound (US) breast images, compared with background areas, tumor is more salient because of its higher contrast. In this paper, we develop a novel automatic localization method based on HVM for automatic segmentation of ultrasound (US) breast tumors. First, the input image is smoothed by convolution with a linearly separable Gaussian filter and then subsampled into a 9-layer Gaussian pyramid. Then intensity, blackness ratio, and superpixel contrast features are combined to compute saliency map, in which Winner Take All algorithm is used to localize the most salient region, presenting with a circle on the localized target. Finally the circle is taken as the initial contour of CV level set to finish the extraction of breast tumor. The localization method has been tested on 400 US beast images, among which 378 images have higher saliency than background areas and succeed in localization, with high accuracy 92.00%. The HVM localization method can be used to localize the tumors, combined with this method, CV level set can achieve the fully automatic segmentation of US breast tumors. By combing intensity, blackness ratio and superpixel contrast features, the proposed localization method can successfully avoid the interference caused by background areas with low echo and high intensity. Moreover, multi-object localization of US breast images can be considered in future employment.

Facial expression has lots of applications in human-computer interaction. Although feature extraction and selection have been well studied, the specificity of each expression variation is not fully explored in state-of-the-art works. In this work, the problem of multiclass expression recognition is converted into triplet-wise expression recognition. For each expression triplet, a new feature optimization model based on Action Unit (AU) weighting and patch weight optimization is proposed to represent the specificity of the expression triplet. Sparse representation based approach is then proposed to detect the active AUs of testing sample for better generalization. The algorithm achieved competitive accuracies of 89.67% and 94.09% for Jaffe and CK+ databases, respectively. Better cross-database performance has also been observed.

In aerial image object detection, small targets present significant challenges due to limited pixel information, complex backgrounds, and sensitivity to bounding box perturbations. To tackle these issues, we propose SO-RTDETR for small object detection. The model introduces a Cross-Scale Feature Fusion with S2 (S2-CCFF) module, a Parallelized Patch-Aware attention (PPA) module, and the Normalized Wasserstein Distance (NWD) loss function, leading to significant performance improvements. Specifically, the S2-CCFF module enhances small object information by incorporating an additional S2 layer, while SPDConv downsampling maintains key details and reduces computational cost. The CSPOK-Fusion mechanism integrates global, local, and large branch features, capturing multi-scale representations and effectively mitigating interference from complex backgrounds and occlusions, thereby enhancing the spatial representation of features across scales. The PPA module, embedded in the Backbone network, leverages multi-level feature fusion and attention mechanisms to retain and strengthen small object features, addressing the issue of information loss. The NWD loss function, by focusing on the relative positioning and shape differences of bounding boxes, increases robustness to minor perturbations, enhancing detection accuracy. Experimental results on the VisDrone and NWPU VHR-10 aerial datasets demonstrate that our approach outperforms state-of-the-art detectors.