采用分段主成分和PPI的高光谱影像分类  

Classification of Hyperspectral Images Based on Segmented Principal Components and PPI

作　　者：梁远玲 简季[1] LIANG Yuanling;JIAN Ji(College of Earth Sciences,Chengdu University of Technology,Chengdu 610059,China)

机构地区：[1]成都理工大学地球科学学院,成都610059

出　　处：《遥感信息》2020年第1期129-134,共6页Remote Sensing Information

摘　　要：高光谱遥感影像波段多且存在混合像元,特征提取以及端元提取都是高光谱影像分类必不可少的工作,分类方法的选择也是因地适宜。以福建省泉州市德化县下属某一地区的CASI影像为实验数据,基于分段主成分(segmental principal component analysis,SPCA)和纯净像元指数法(pure pixel index,PPI),提出了最小距离(minimum distance classification,MDC)和二进制编码(binary encoding,BE)的高光谱影像分类方法。实验结果表明,MDC的总体精度为69.71%,BE的总体精度为70.88%。对单一地物精度而言2种方法各有其长,MDC对道路的分类精度更高,为98.08%;而植被、耕地和水体采用BE方法的分类精度更高,分别为94.12%、98.08%、98.11%。本文提出的方法应用于CASI高光谱影像,对该研究区的地物分类研究有一定的实用性和参考价值。Hyperspectral remote sensing images have multiple bands and mixed pixels.Both feature band extraction and endmember extraction are essential for hyperspectral image classification,and the selection of classification method is also suitable for the locality.In this paper,the images of a region subordinated to Dehua county,Quanzhou city,Fujian province,are selected as experimental data.Based on segmental principal component analysis(SPCA)and pure pixel index(PPI),a classification method of hyperspectral images using minimum distance classification(MDC)and binary encoding(BE)is proposed.The experimental results show that the overall accuracy of MDC is 69.71%and that of BE is 70.88%.Both methods have their advantages for the accuracy of a single ground object.MDC method has a higher classification accuracy of 98.08%for roads,and the classification accuracy of BE method for vegetation,cultivated land and water body is higher,which is 94.12%,98.08%and 98.11%respectively.The proposed method is applied to the CASI hyperspectral imaging,which is of practical and referential value for the classification study of the ground objects in the research area.

关 键 词：分段主成分分析 纯净像元指数法 最小距离法 二进制编码 高光谱分类 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]