This paper is the result of geological survey engineering.[Objective] Water conservancy projects are usually located in mountainous areas with steep terrain, inconvenient transportation and harsh environment, so that the survey work is very difficult. Meanwhile, the limitation of horizon makes it impossible for geologists to grasp the features of geological units as a whole, especially in high mountain and canyon areas. The application of hyperspectral remote sensing technology in the geological field mainly focuses on the extraction and identification of geological composition. The types of surface features are determined by similarity of spectral features between unknown geological units and known mineralsand rocks. Using its extremely high spectral resolution, spatial resolution and the advantage of union of imagery and spectrum, the technology can be applied on fine classification of rock types in alpine and canyon areas, so as to improve the efficiency of geological survey and reduce costs and risks. [Methods] This paper introduces Bei Fang Investigation, Design & Research Co., Ltd. adopting the SSMAP-II small short-wave infrared imaging system independently developed by the Nanjing Center, China Geological Survey in the project of "Research on Remote Sensing Geological Interpretation Methods Based on Hyperspectral Image Data", taking a helicopter as platform to carry out hyperspectral ground-based remote sensing surveys in Diebu County, Gannan Tibetan Autonomous Prefecture. The surveys aerial hyperspectral image data of the study area, then used MNF minimum noise fraction(MNF) method and band ratio analysis method to interpret the lithology. [Results] The interpretation result is basically consistent with the lithology distribution of the known regional geological map, providing first-hand information for the water conservancy projects of the dangerous areas. [Conclusions] This research improves the work efficiency and the level of geological exploration, meanwhile it also verifies the reliability and practicability of the domestic hyperspectral imaging spectrometer.