【摘要】 针对高光谱图像中背景及目标先验知识未知条件下的异常目标检测问题,提出了一种基于独立成分分析(ICA)的异常探测算法。首先估计原始数据的虚拟维(VD)以确定要分离的独立成分个数,在此基础上进行快速独立成分分析(FastICA),然后基于平均局部奇异度选择含异常信息较多的独立成分,最后使用丰度量化算法得到异常目标的丰度图像。为了验证算法的有效性,对由AVIRIS获取的真实高光谱图像进行了异常检测实验,并与经典RX算法和LPD算法的检测结果进行了比较。结果表明,基于ICA的检测算法具有良好的检测性能和较低的虚警,且运算复杂度较低。更多还原

【Abstract】 Based on independent component analysis(ICA) anomaly detection algorithm is proposed to deal with detecting unknown targets in unknown background for hyperspectral imagery.First,virtual dimensionality(VD) is introduced to determine the number of independent components required to be generated by FastICA.Then,the independent component which has the most information about anomaly targets is selected based on its average local singularity.Finally,an ICA-based abundance quantification algorithm is applied to produce the abundance fraction map of the anomaly targets.A real AVIRIS hyperspectral data set is tested for anomaly detection.The experimental results show,the proposed method outperforms RX and LPD,has lower false alarm probability and lower computational complexity.更多还原