一种多/高光谱遥感图像端元提取的凸锥分析算法

A Convex Cone Analysis Method for Endmember Selection of Multispectral and Hyperspectral Images

凸锥分析方法常用于多光谱和高光谱遥感图像的端元提取。遥感图像中的每个像元都可以看作一个多维向量,整幅影像看作由离散的非负向量构成的凸锥,通过寻找凸锥的角点来自动获取图像的端元。本文提出了一种自动选择最佳凸锥角点的方法,应用到传统的凸锥分析方法中,提高了凸锥分析方法的效率。利用模拟数据和真实数据实验验证了算法的可行性。

Convex Cone Analysis（CCA） method can be applied to endmember selection from multispectral and hyperspectral imagery. Each pixel on multispectral and hyperspectral imagery can also be regarded as one vector and the whole image is a convex cone formed by a number of nonnegative discrete vectors, so endmember selection is equivalent to search for the vertices of a convex cone, A method of automatically selecting best corners（vertices） is presented, which improves the traditional CCA method, Experiments on simulated data and real data verify the validity of CCA method.