摘要
针对高光谱图像混合像元影响异常检测效果的问题,提出了一种基于端元提取的异常检测算法。该算法采用小波分解,将原始高光谱图像分解为高频信息图像和低频信息图像,舍弃低频信息图像,只利用高频信息图像,从而抑制了背景,突出了目标;然后使用正交子空间投影(OSP)方法提取图像的端元光谱;最后根据提取的端元光谱,采用光谱角匹配(SAM)技术完成高光谱图像的异常检测。为了验证本文方法的有效性,利用AVIRIS高光谱数据进行了仿真实验,取得了较好的检测效果。与其他算法相比,结果表明,本文算法的检测性能明显优于传统算法,既降低了虚警率,又大大缩短了计算时间,适用于实时的高光谱图像异常目标检测。
In order to overcome the bad influence caused by mixed pixels in hyperspectral anomaly detection, a new target detection algorithm based on endmember extraction method is proposed. Hyperspectral imagery is decomposed into high frequency and low frequency images by wavelet decomposition firstly. High frequency images used only and sight of low frequency images abandoned, background information is effectively suppressed and anomaly targets become obvious consequently. And then the endmember spectral profile is got from high frequency images by Orthogonal Subspace Projection（OSP） algorithm. At last, anomaly detection is done by Spectral Angle Mapping（SAM） in the light of extracted endmember spectra. The proposed algorithm is studied using real hyperspectral data, and good detection effect is obtained. The results show that the proposed method which needs less time and has lower false alarm rate is proved to be better than the traditional algorithm, thus it is suitable for real-time anomaly detection in hyperspectral imagery.
出处
《红外技术》
CSCD
北大核心
2015年第10期836-841,共6页
Infrared Technology
基金
吉林省科技发展计划资助项目
编号:20140101213JC
全军军事学研究生课题项目
编号:2011JY002-534
关键词
高光谱图像
异常检测
端元提取
小波分解
正交子空间投影
hyperspectral imagery
anomaly detection
endmember extraction
wavelet decomposition
orthogonal subspace projection
