As an very important part of the Qinghai-Tibet Plateau ecosystem, it is of great significance to study alpine shrubs. But for a long time, due to the remote location and underdeveloped transportation, as well as the harsh growing conditions, the alpine shrub on the Qinghai-Tibet Plateau has been less studied. Remote sensing detection technology can overcome the difficulties caused by geography and environment and can be used to detect large areas and non-destructive. Therefore, remote sensing detection technology can be used to study alpine shrubs in Qinghai-Tibet Plateau. As the traditional high-resolution remote sensing detection technology is often adopted with three bands of RGB, the discrimination accuracy of different plants is low, and the difference of NDVI′ index and RVI′ index of corresponding plants is small, which cannot effectively distinguish various types of vegetation. At the same time, hyperspectral reflectance curve and irradiance curve contain spectral information of thousands of bands. If a single band is selected for plants detection, the loss of spectral information is very large, and the characteristics of thickets reflected are not obvious, resulting in low confidence. In order to distinguish the alpine shrub vegetation, this paper uses hyperspectral technology to carry out spectral characteristic analysis of the shrub, providing theoretical support for remote sensing detection of the shrub on the Qinghai-TibetPlateau. The research draws support from FieldSpec4 high resolution spectrometer of the America. It was used to identify 6 shrubs (Rhododendron capitatum, Caraganajubata, Potentillafruticosa, Salix cupularis, Daphne tangutica and Berberisdiaphana) grown in the eastern Qilian Mountains through measuring the reflectance rate and absorption rate, calculating the first order differential of absorption rate (GREF and GABS) to enlarge the resolution of spectral curve, screening the sensitive wavelength, and then identifying different shrubs by calculating their values with NDVI and RVI. The result indicated that (1) the absorption spectral curves of shrubs were similar with most plants, but their first absorption valley shifted to left; (2) the shrubs performed unique spectral features in some sensitive wavelengths, and these features could be used to improve the resolution by REF, ABS, GREF and GABS transformation to identify the shrublands; (3) The spectral values of the 6 shrubs are different, and the relatively stable wavelengths are 550～680, 860～1 075, 1 375～1 600 and 1 900～2 400 nm. Therefore, these 4 wavelengths can be selected as sensitive areas to identify shrub plants; (4) NDVI and RVIcalculated with the REF average value of sensitive wavelengths of 575～673 and 874～920 nm and/or the area value of sensitive wavelengths of 685～765, 556～590, 635～671 and 1 117～1 164 nm could effectively identify 6 shrubs.