A Comparison of Four Algorithms for Land-Use Classification Based on Landsat 8 OLI Image

Abstract

Accurate mapping and monitoring of land-use is essential for reasonable land management and planning. To extract land-use classes based on remote sensing images, many classification algorithms have been proposed. However, the comparison between some main supervised classification algorithms is rarely researched. This study selected the eastern fringe area of Jinan as the study area and the Landsat 8 OLI image of 2019 as data to compare the performance of four supervised classification algorithms that are MLC, SVM, ANN and RF especially. The results shown that the overall accuracy and kappa of RF is 86.2% and 0.8, and the overall accuracy and kappa of SVM is 83.2% and 0.75, and the overall accuracy and kappa of ANN is 81% and 0.72. The overall accuracy and kappa of MLC is 73.6% and 0.63. These denote that the RF can achieve the best classification result in four algorithms, followed by SVM, ANN and MLC.