The Study Of Identification And Research Method For Hyperspectral Mixed Pixel Based On Modified Gaussian Model

Since the coming of concept of imaging spectrometer, the hyperspectral remote sensing is always on the cutting-edge of remote sensing science technologies, and has been developed constantly. Comparing with the traditional multi-spectral remote sensing image, the hyperspectral image is a kind of rich-information and high-resolution spectra image. Especially, this technology has broad applied in the remote sensing geological field, and well used in the geological exploration and geological prospecting prominently. That makes it possible that using the hyperspectral remote sensing image information on identifying the objects which is on the surface of the land.Through the research of the spectrum of high-pixel mixed approach and in the view of that the modified gaussian model method is the current hot model method in the hyperspectral remote sensing field, this paper applies the method to the identifying the component of the mixed hyperspectral image pixels, which uses this method to achieve the realization of mixed-pixel element of the purpose of identification. The modified gaussian model method uses a series of modified gaussion curves to fit the real spectral curves, and the comparing the parameters of each Gaussian curves to the parameters, which comes from the mineral samples standard spectra library. And then, according the comparing result, classify the species of the mixed pixel. The method receives a good result in the analyse of hyperspectral image mixed pixel.This article choose alunite, chlorite, goethite, hematite, illite and kaolinite six mineral samples as samples library. Though the test of some mixed pixels which are chose from a real hyperspectral image, the test result achieve the basic purpose of the experiment, so as to provide feasibility of the modified gaussian model method.