将传统遥感图像分类方法中的光谱角度制图法(Spectral Angle Mapping-SAM)加以变换,改进为一种符合全约束条件下的高光谱遥感图像的混合像元分解模型.新算法在端元丰度比例满足全约束的条件下,通过逼近的方法寻找一种端元丰度的比例组合,使测试光谱与目标光谱的广义夹角最小,从而认为该比例组合就是混合像元分解的结果.试验结合高光谱遥感模拟图像进行了分解实验,同时与最小二乘法做了比较,结果表明,新算法不仅严格地将各种端元组分的丰度值控制在0到1之间,而且其分解结果与模拟图像的实际情况也比较吻合,总体上新算法要优于最小二乘法.%It is an important prerequisite in sub-pixel mapping of hyperspectral remote sensing image, that mixed pixels must be effectively decomposed under all the constraints. In this paper, as one of the traditional image classification methods, the Spectral Angle Mapping method(SAM)was transformed and improved.so a new mixed pixel decomposition model which is under all constraints was proposed. Under the condition that the proportions of all endmembers meet all constraints, the new algorithm tries to find a kind of combination of the endmember proportion by approximation method, which can make the angle between test spectrum and target spectrum minimum, and the proportions of endmember are to be as the result of pixel decomposition. The new algorithm was tested on simulated hyperspectral data, and the result shows that the new algorithm works very well, the proportions of all endmembers not only are strictly controlled to be between 0 and 1, but also are more consistent with simulated hyperspectral data. In general, the new algorithm is superior to the least square method.
展开▼
