A Comparative Study on Methods of Extracting Land Cover Informations Based on Landsat 8 In Dianchi Basin

Abstract

Using remote sensing software ENVI5.1, combined with Landsat 8 data, the land cover information of the Dianchi Lake Basin is classified and extracted by maximum likelihood classification of supervised classification, ISODATA algorithm of unsupervised classification, and decision tree classification.The classification results and classification accuracy were obtained. Accuracy evaluation and comparative analysis of each classification method.The results show that the overall accuracy of supervised classification in the land cover classification in the study area is 93.90%, the overall accuracy of unsupervised classification is 85.72%, and the overall accuracy of decision tree classification is 75.59%.The supervised classification accuracy is higher than that unsupervised classification and decision tree classification.The categories extracted by supervised classification are continuous and the boundaries are clear, and supervised classification effect is basically consistent with the actual situation. Among them, the accuracy of the producers of forest land, agricultural land, build land, unused land and water area has reached more than 90%.