Abstract: Taking Landsat 8 OLI Satellite images and DEM of Xishuangbanna in 2015 as basic data. Based on the pixel-based classification method, the best index of forest extraction was screened out through comparative analysis of experiments.Using a variety of landscape indices, the degree of fragmentation of the tropical rainforest in the elevation, aspect and slope of Xishuangbanna was analyzed. The results showed that NDVI could effectively extract accurate spatial information of tropical rainforests and the accuracy was 95.66%. The forest coverage of Xishuangbanna was 77.8% in 2015. Forest distribution in elevation and slope appeared a skewed distribution, and was uniform in slope aspect. The topographic differences of forest landscape fragmentation were obvious. Forest landscape fragmentation was the most serious below 500 m and in plain. With the increases of elevation and slope, PD and AI continuously decreased and increased, forest landscape fragmentation gradually reduced. PD was highest and AI was the lowest at North Slope Aspect, this illustrated forest landscape fragmentation was obviously higher than other slope aspects.