The Qinghai-Tibet Plateau is a hotspot area for global climate change research and a key area for ecological protection in China. Land cover change research in this area can significantly contribute to optimizingland use pattern and improving ecological services and natural conditions of the region. Among various methods for land cover change detection, linear spectral unmixing (LSU) is an effective approach to monitor land cover change by using remote sensing technology. In this study, we chose Naidong County in Tibet as the study area and adopted the linear spectral unmixing technology to detect the ratio of vegetation, bare soil, and rock for each pixel of three Landsat TM/ETM images from 1988, 2000, and 2010. The vegetation component of the result was compared with NDVI. The results show that: (1) The study area has high proportion of mixed pixels, and LSU can deal with the complex features of land cover changes effectively. (2) The proportion of bare soil reduced evidently while rock and vegetation coverage increased from1988 to 2010. It proves that vegetation cover had recovered to some degree while rocky desertification expanded rapidly. (3) Through comparing the vegetation component with NDVI we conclude that LSU is applicable for the identification of land cover change of the Qinghai-Tibet Plateau.