PreprintArticleVersion 1Preserved in Portico This version is not peer-reviewed
Comparing the Performance of Different Classification Algorithms for Mapping and Assessing Land Cover Changes in Areas with Surface Mining and Complex Landscape Using Landsat Imagery
Version 1
: Received: 17 May 2023 / Approved: 18 May 2023 / Online: 18 May 2023 (12:39:51 CEST)
How to cite:
Vorovencii, I. Comparing the Performance of Different Classification Algorithms for Mapping and Assessing Land Cover Changes in Areas with Surface Mining and Complex Landscape Using Landsat Imagery. Preprints2023, 2023051345. https://doi.org/10.20944/preprints202305.1345.v1
Vorovencii, I. Comparing the Performance of Different Classification Algorithms for Mapping and Assessing Land Cover Changes in Areas with Surface Mining and Complex Landscape Using Landsat Imagery. Preprints 2023, 2023051345. https://doi.org/10.20944/preprints202305.1345.v1
Vorovencii, I. Comparing the Performance of Different Classification Algorithms for Mapping and Assessing Land Cover Changes in Areas with Surface Mining and Complex Landscape Using Landsat Imagery. Preprints2023, 2023051345. https://doi.org/10.20944/preprints202305.1345.v1
APA Style
Vorovencii, I. (2023). Comparing the Performance of Different Classification Algorithms for Mapping and Assessing Land Cover Changes in Areas with Surface Mining and Complex Landscape Using Landsat Imagery. Preprints. https://doi.org/10.20944/preprints202305.1345.v1
Chicago/Turabian Style
Vorovencii, I. 2023 "Comparing the Performance of Different Classification Algorithms for Mapping and Assessing Land Cover Changes in Areas with Surface Mining and Complex Landscape Using Landsat Imagery" Preprints. https://doi.org/10.20944/preprints202305.1345.v1
Abstract
In order to conduct an accurate classification of the heterogeneous landscape in Jiului Valley, Romania mining basing, four machine learning algorithms (SVMs) and two common algorithms (MLC and MD) have been compared, using a temporal series of Landsat satellite images from the period 1988-2017. By using independent validation, an accuracy assessment was established together with the analysis of the differences between the classification algorithms used. Although all six algorithms used have shown a high overall accuracy (ranging from 80.29% to 93.14%) and Kappa values (from 0.77 to 0.92), SVM-RBF appears to have a higher overall applicability in describing the spatial distribution and the cover density of each land cover category. Results have indicated a large difference in classification accuracy between the SVM-RBF algorithm and commonly used algorithms, the SVM-RBF algorithms have slightly outperformed the MLC with an overall accuracy of 7.14–8.86% and by 0.0833–0.1033 kappa coefficient. On the other hand, the same algorithm have outperformed the MD by and overall accuracy of 9.71–10.86% and by 0.1133–0.1267 kappa coefficient. By using SVM-RBF, certain classified maps have been developed and used for assessing changes by post classification comparison. The results have shown an average growth of 6.5% in mined areas over the studied period.
Keywords
Support Vector Machine; Maximum Likelihood; Minimum Distance; machine learning; classification algorithm; Landsat
Subject
Environmental and Earth Sciences, Remote Sensing
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.