将n个波段的高光谱图像像元与n维空间里的散点联系起来，结合凸体几何中单形体概念研究高光谱遥感图像纯净像元提取方法，实现图像的地物精确分类识别及像元波谱分解。寻找高光谱遥感图像n维空间里的单形体并认知分析单形体是该研究方法的重要环节。通过MNF(minimum noise fraction)变换和PPI(pixel purity index)计算技术寻找到单形体，基于单形体进行像元分解分析单形体，并结合应用实例和SAM(spectral angle mapper)分类技术完成高光谱图像地物精确分类制图，验证了该研究方法的可操作性。该研究方法的优点在于不需要用户提供地物波谱信息，用于制图和波谱分解的终端单元可由图像本身得到，并由用户控制分类制图和波谱分解的详细程度。

One advantage of hyperspectral remote sensing is that it has more bands so more information could be used to recognize ground objects and estimate relative contents of materials. In this paper, pixels of hyperspectral remote sensing images of n bands are connected with points in an n-dimensional scatterplot. Pure pixels can be extracted using a method of simplex, which is a concept in convex geometry, and thus accurate hyperspectral image classification and spectral unmixing can be realized. The focus of this method is to find the simplex and to analyze it. The simplex can be found using MNF(minimum noise fraction) transform and PPI(pixel purity index) calculation, and the mapping methods used here are SAM(spectral angle mapper) classification and an unmixing method based on the simplex. All techniques here have been proved feasible by an application example. This paper also gives a procedure of the techniques. The advantages of the techniques and the procedure are that the endmenmbers for spectral mapping and unmixing can be extracted from the images themselves, and that spectral mapping and unmixing scale can be determined by users.