In this paper, Baogutu No.5 rock mass of Westtern Junggar in Xinjiang is taken as an example, wall rock alteration zoning and periodical pattern are systematically discussed, combined with ETM+ data, and with the help of remote sensing images pre-processing methods, a multi-spectral remote sensing alteration information extraction model based on spectral fitting technology of typical alteration minerals has been set up, namely CES model composed by remote sensing data calibration, mask, pixel unmixing, spectral fitting, SAM, Classification. Based on this model, information extraction has been conducted on the rock mass, and has been overlapped with geochemical data through which feasibility of this method has been validated, satisfactory results have been achieved.