摘要
提出了一种利用独立成分分析(ICA)正交子空间投影加权的高光谱影像目标探测方法。该方法从影像像元集合的独立成分入手,通过一种光谱相似性测度加权,赋予每个像素合适的权值,从而有效地解决从原始影像中无法正确提取背景数据而造成的虚警概率高的问题。实验结果表明,相比于经典的CEM方法,在相同的探测概率下,该方法能降低1.97%的虚警概率;与相关目标探测算法相比,所提出的算法具有较好的目标探测效果。
Hyperspectral data contained over hundreds of narrow contiguous wavelength bands are extremely suitable for target detection due to their high spectral resolution. In the target detection for hyperspectral image, the background data are not well represented from the original data sources. We propose a weighted hyperspectral image target detection algo- rithm based on independent component analysis orthogonal subspace projection（ICA-OSP）. The methods start from a collection of independent component of the image pixels, through a spectral similarity measure weighted so that each pixel to give the appropriate weights. It can effectively solve the problems that can not correctly extract the background data from the o- riginal image. The problem usually causes a higher false alarm probability. AVIRIS hyper- spectral image simulation and detection algorithms are compared by ROC curves with the rel evant target detection algorithm, and the results show that the proposed algorithm can re- duce the false alarm probability, to better target detection effects.
出处
《武汉大学学报（信息科学版）》
EI
CSCD
北大核心
2013年第4期440-444,共5页
Geomatics and Information Science of Wuhan University
基金
国家973计划资助项目(2006CB701303)
中央高校基本科研业务费专项资金资助项目(20102130202007)
关键词
独立成分分析
正交子空间投影
光谱相似性测度加权
高光谱影像
independent component analysis
orthogonal subspace projection
spectral simi-larity measure weighted
hyperspeetral image
