一种遥感影像裸土地特征增强方法

摘要

It's difficult for feature enhancement algorithm available to highlight bare land in a given band. It presents a new feature enhancement method for bare land in multi-spectral remote sensing images. The image is partitioned coarsely by spectral distance, and bare land and non-bare-land pixel samples are obtained. The image is transformed with the whitening matrix form non-bare-land pixel samples. The whitened image is rotated with the principal components transformation matrix from its bare land pixel samples. One procedure for doing this is discussed, to validate the method, an example is presented.%针对常用特征增强算法难以将裸土地在特定波段突出出来的问题,提出了一种适用于多光谱遥感影像的裸土地特征增强算法.基本思路是采用根据光谱距离进行预分割,估计裸土地和非裸土地样本,针对非裸土地样本进行白化处理,基于白化数据中的裸土地样本进行主分量变换,裸土地信息将被集中并突出于变换结果的第一分量.对算法的具体流程进行了探讨,并通过一个实例进行了验证.