Abstract: The classification of land desertification is an important part of desertification monitoring, and it is also the basis for comprehensive management and scientific protection of land desertification. Aiming at the problem of abnormal extraction of land desertification in arid areas, this paper selects the arid/semi-arid Horqin area as the experimental area. The medium-and high-resolution Landsat remote sensing images of 2005, 2010 and 2015 are used as the data sources. Based on a large number of statistical analysis of samples, a desertification extraction model that integrates vegetation coverage(FVC), modified soil adjusted vegetation index(MSAVI) and the enhanced vegetation index(EVI) is put foward, which is compared with the extraction results of traditional vegetation coverage indicators. The research results show that compared with the single vegetation index retrieval method, the algorithm proposed in this paper has higher classification accuracy, especially for arid/semi-arid areas, the method of fusion vegetation index has better applicability and robustness. This method provides new ideas for the establishment of the desertification evaluation system, and provides auxiliary decision support for the protection and management of land desertification.

Key words: desertification, vegetation coverage, vegetation index, Landsat, decision tree classification