基于新型相似性测度的高光谱影像地物检测  

A Novel Similarity Measurement Based on Object Detection in Hyperspectral Imagery

作　　者：陈伟[1] 余旭初[1] 张钢[2] 张立福[3] 张鹏强[1] 

机构地区：[1]信息工程大学测绘学院 [2]69028部队 [3]中国科学院遥感应用研究所遥感科学国家重点实验室

出　　处：《测绘科学技术学报》2012年第1期42-46,共5页Journal of Geomatics Science and Technology

基　　金：国家自然科学基金资助项目(41072248)

摘　　要：提出了一种新型光谱相似性测度及其参数的自适应选择方法,并且将其应用到了高光谱影像地物检测中。由于这种相似性测度基于光谱角度余弦(SAC),因此在理论上对因光照强度变化、阴影和遮挡等引起的同种地物光谱变化的适应性较强。最后利用两幅高光谱影像进行了实验分析,实验结果证明提出的方法不仅能扩大阈值取值区间,而且可提高检测的精度。A novel spectral similarity measurement and its adaptive selection of parameter were proposed and applied to object detection in hyperspectral imagery.As the novel similarity measurement is based on spectral angel cosine（SAC）,in theory,it has a good adjustability for spectral curve variation for the same materials coming from radiation intensity variations,shadow,and shading etc.Lastly,the experiments were carried on with two hyperspectral images,and the results of the experiments proved that the proposed method could not only extend the threshold area coverage but also improve the precision of detection.

关 键 词：高光谱影像 地物检测 光谱相似性测度 参数自适应选择 光谱角度余弦 

分 类 号：P237[天文地球—摄影测量与遥感]