一种基于PCA-LDA的卫星遥感图像的分类方法

摘要

We propose a colour classification algorithm for satellite remote sensing images based on PCA-LDA, which integrates the features of components analysis (PCA) and linear discriminant analysis (LDA). The algorithm fuses the feature spaces of PCA and LDA algorithms to get the colour-fused feature space. The algorithm removes the correlation between R, G and B of the image, improves the light sensitivity and classifies the colours of the image using region classification-based spatial consistency principle. Experimental results demonstrate that the PCA-LDA-based algorithm is an effective method to classify the remote multi-frequency sensing image.%结合主成分分析PCA(Principal Components Analysis)和线性判别分析LDA(Linear Discriminant Analysis)的特点,提出一种基于PCA-LDA算法的卫星遥感图像色彩分类方法.该算法将PCA算法和LDA算法的特征空间相融合,得到融合颜色特征空间.该方法去除了图像的R、G、B之间的相关性,改善了光照敏感性,采用基于区域分类的空间一致性原则对图像进行颜色分类.实验结果表明,该方法是对多频谱遥感图像分类的一种有效的方法.