基于地物信息的高光谱遥感图像分类方法

Classification Method of Hyperspectral Remote Sensing Image Based on Feature Information

利用主成分分析法滤除n维高光谱遥感图像中的大部分冗余信息,得到尽可能保留光谱信息的m维高光谱遥感图像,融合其地物空间分布信息,将m维高光谱遥感图像中的每一个像素点构建为一个2维谱-空信息向量。再利用主成分分析法法对m维高光谱遥感图像进行降维,得到q维融合地物空间分布信息与光谱信息的结果图。通过高斯混合模型预测聚类中心,基于改进的迭代自组织数据分析算法ISODATA(Iterative Selforganizing Data Analysis Techniques Algorithm)对高光谱遥感图像进行聚类,得到最终的分类结果。实验结果表明本方法的地物分类精度优于K⁃means、ISODATA和SVM方法,总体分类精度提升10.14%-13.99%,kappa系数提升3.2%-12.85%。

principal component analysis to filter out most of the redundant information in the n⁃dimensional hyperspectral remote sensing image is used,and the m⁃dimensional hyperspectral remote sensing image that retains the spectral information is obtained as much as possible its spatial distribution information,each pixel in the m⁃dimensional hyperspectral remote sensing image is constructed as a 2⁃dimensional spectrum⁃space information vector.The principal component analysis method is used to reduce the dimension of the m⁃dimensional hyperspectral remote sensing image again,and the result map of the q⁃dimensional fusion of the spatial distribution information of the ground features and the spectral information is obtained.The Gaussian mixture model is used to predict the clustering center,and the hyperspectral remote sensing image is clustered based on the improved ISODATA(Iterative Selforganizing Data Analysis Techniques Algorithm)to obtain the final classification result.Experimental results show that the classification accuracy of the proposed method is better than that of K⁃means,ISODATA and SVM,the overall classification accuracy is increased by 10.14%-13.99%,and the kappa coefficient is increased by 3.2%-12.85%.