利用小波分解和顶点成分分析的高光谱异常检测

针对高光谱图像复杂背景影响异常检测结果的问题,提出了一种新的抑制背景的异常检测算法。首先对高光谱图像采用小波分解,将高光谱图像分解成高频图像和低频图像;然后使用顶点成分分析(VCA)方法提取高频图像的端元光谱图;最后使用光谱角匹配(SAM)技术对高光谱图像进行异常检测。实验结果表明,KRX算法相比,本文算法像素目标个数增加32.35%;虚拟个数减少12.12%,计算时间少2个数量级;本文方法的ROC曲线一直位于KRX、PCA-KPX算法的异常检测方法之上,有利于高光谱图像的实时异常检测应用。

Anomaly detection algorithm based on wavelet decomposition and vertex component analysis in hyperspectral images

In order to overcome the bad influence caused by complex background in hyperspec tral images a new approach for anomaly detection based or wavelet decomposition and vertex component analysis is proposed.Hyperspectral data is decomposed by wavelet decomposition firstly into high frequency images and low frequency image.And then the endmember spectral profile is got from high frequency images by vertex component analysis.At last,anomaly detection is do ne by spectral angle mapping.The method which needs much less time is proved to be better than the KRX and PCA-KRX algorithms by ROC,and compared with KRX algorithm,the target pixel s obtained by the proposed idea are increased by 32.35%,but the false alarm pixels are decreas ed by 12.12%.