线性混合光谱模型在多光谱遥感影像分类中的应用

Application of Linear Mixing Spectral Model to Classification of Multi-spectral Remote Sensing Image

介绍了混合像元的概念、研究现状和广泛用于混合像元分解的线性模型。以TM遥感图像分类为例,对图像做MNF处理后进行有约束和无约束两种线性混合光谱模型分解混合像元;再对图像进行分类,将得到的图像同直接分类的图像进行比较分析。实验结果表明,考虑约束条件的线性混合光谱解混分类精度比直接分类精度提高了32.8%。

The concept of mixed pixel, development situation, and the linear mixing spectral model and their resolution method were briefly introduced. Taking the processing of TM image as an example, the MNF translation was made and the mixed pixels were unmixed with linear mixing spectral models including restricted and non-restricted models. And then the images were classified and compared with the firsthand classification without processing. Result indicates that the restricted resolution method increases the accuracy by 32.8%