Abstract: Mountain vegetation information plays an important role in climate change research and ecological environment protection. Remote sensing technology can quickly acquire mountain vegetation information, but there are the influence of mountain terrain shadows and mountain vegetation information confusion. This paper took mountain vegetation as the research object and analyzed the main characteristics of mountain vegetation based on the multi-spectral data of Landsat satellite remote sensing image. Learn from the structural principle and form of the shadow elimination vegetation index (SEVI), a vegetation index algorithm - Vegetation distinguished and shadow eliminated vegetation index (VDSEVI) for mountain vegetation cover remote sensing monitoring was proposed. Samples for comparison and analysis were selected according to the main land cover types in the study area. The accuracy, validity and practicability of mountain vegetation information extraction with different vegetation indices were compared and analyzed. There are certain criteria which vegetation index of the same vegetation cover in shady and sunny should be equivalent, and the vegetation index values of different vegetation cover should be differentiated and compliance with actual vegetation coverage. Comparative analysis methods for different vegetation indices include: the images of different vegetation indices were directly compared; the vegetation index values of the same vegetation in shady and sunny cover were compared; the vegetation index values of the different land cover types were compared; the correlation between vegetation index and cosi was analyzed. The VDSEVI was compared with the ratio vegetation index(RVI), the normalized vegetation index(NDVI), the enhanced vegetation index(EVI2) and SEVI. There was a significant difference in the mean value of VDSEVI among different land cover types. The relative error of the sparse woodland in shady and sunny was small, which was 3.428%. The standard deviation of each land cover type sample was less than 0.060.The shady area of the Nalati in Xinjiang was dominated by woodland, and the sunny area was dominated by grassland. Therefore, the vegetation index and cosi should be negatively correlated. The correlation coefficient between VDSEVI and cosi is -0.800.The data results showed that compared with RVI, NDVI, EVI2 and SEVI, VDSEVI eliminated the influence of terrain shadows, and had a large amount of information and strong recognition of vegetation coverage. The problem of vegetation information confusion was solved, and the actual situation of mountain vegetation cover was reflected. To verify the suitability of VDSEVI in other regions, VDSEVI was applied to the area of Arxan in Inner Mongolia and Minhou County in Fuzhou City. The results showed that VDSEVI was equally effective. Vegetation information was extracted based on VDSEVI threshold method in Nalati in Xinjiang, Arxan in Inner Mongolia and Minhou County in Fuzhou City. The threshold was calculated by the VDSEVI value of samples from comparative analysis of various land cover types. Finally, the results of vegetation information extraction were evaluated by validating samples. The overall accuracy of vegetation information extraction in the three regions was 84.136%, 87.339%, 86.709% respectively, and the Kappa coefficients were 0.799, 0.788 and 0.791 respectively.