基于ICA的遥感图像的色彩分类方法

Classification of Remote Sensing Image Based on Independent Components Analysis

根据独立成分分析(ICA)方法和多频谱卫星遥感图像的特点,提出了一种基于ICA的遥感图像色彩分类法。方法使用Fast ICA算法提取遥感图像的色彩独立成分,是RGB反转的结合,具有互补的分布,不受照明的影响。使用最大相似度分类算法对像素进行色彩分类,实验结果表明,方法的色彩分类效果较好,对多频谱遥感图像进行色彩分类十分有效。

This article propose a classification algorithm for satellite remote sensing images based on Inde- pendent Components Analysis（ICA）. The algorithm combines the advantage of ICA and muhispectral re- motely sensed images. The algorithm extracts the spectral independent components of multispectral re- , motely sensed images by Fast ICA algorithm. It is the combine of reversion about R, G and B, has comple- mentary distribution and is unacted on illumination. Maximum Likelihood is used to classify the pixels. Experimental results demonstrate that the algorithm is an effective improve method to classify the multi- spectral remotely sensed images.