基于支持向量回归的高光谱影像目标探测  

Target detection in hyperspectral imagery using support vector regression

作　　者：陈伟[1] 余旭初[1] 王鹤 

机构地区：[1]信息工程大学测绘学院,郑州450052 [2]北京望神州科技有限公司,北京100020

出　　处：《测绘科学》2010年第3期156-158,共3页Science of Surveying and Mapping

基　　金：信息工程大学测绘学院学位创新创优基金的支持

摘　　要：高光谱影像目标探测可视为一个分类问题,本文通过揭示支持向量回归(SVR)与支持向量分类(SVC)之间的关系,证明了SVR用于分类的可行性,并以此为根据提出了一种基于SVR的目标探测算法,该算法利用虚拟维数得到端元个数的估计,结合端元选择和线性混合模型生成训练样本替代从影像中选择的训练样本,因而减少了对影像先验知识的依赖。采用模拟数据和由AVIRIS获得的高光谱影像对本文算法进行了检验,结果令人满意。Target detection in hyperspectral imagery could be regarded as a classification subject. This article revealed the relationship between support vector regression （SVR） and support vector classification （SVC）, and proved that SVR could be used in classification, as well as brought forward a target detection algorithm based on SVR. The algorithm used virtual dimensionality to estimate the number of endmember, and combines endmember extraction and linear mixing model to generate train samples, which could replace the train samples that came from images. So the dependence on a priori knowledge about the images could be reduced greatly. The paper carried out the experiments by simulative hyperspectral data and actual hyperspectral imagery collected by AVIR/S, and the result of the experiment was satisfactory.

关 键 词：高光谱影像 支持向量分类 支持向量回归 目标探测 线性混合模型 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]