基于空间像素纯度指数的端元提取算法

Endmember extraction algorithm based on spatial pixel purity index

为了减小光谱变化以及异常像素点对端元提取结果的影响,根据局部区域内纯像元和混合像元光谱特征的不同,提出一种基于空间像素纯度指数的端元提取算法.将光谱角距离和欧氏距离加权相加作为新的混合距离测度;采用固定大小的邻域窗口计算图像中所有像素的空间像素纯度指数,在此基础上,根据光谱角距离测度和设定的端元光谱区分性阈值依次搜索端元.仿真数据和真实高光谱图像实验结果表明:该算法能够准确地提取图像中的端元,并且精度高于其他一些端元提取算法.

In order to reduce the effects of spectral variability and existence, of anomalous pixels on endmember extraction results, an endmember extraction algorithm based on spatial pixel purity index was proposed, according to the different spectral characteristics of pure pixels and mixed pixels in local areas. The weighted addition of spectral angle distance and Euclidean distance was utilized as a new mixed distance metric, and then a fixed-size neighboring window was applied to compute the spatial pixel purity index of all the pixels in the image. On this basis, the endmemebers were searched sequentially based on the spec- tral angle distance metric and the predefined discrimination threshold of endmember spectra. Experimental results on both synthetic and real hyperspectral images demonstrate that the proposed algorithm can ex- tract endmembers accurately, and it outperforms several other popular algorithms.