论文部分内容阅读
针对异常检测算法检测精度远低于目标检测算法的问题,提出一种基于近似目标后验信息的高光谱异常检测算法.该算法首先利用基于低秩和稀疏矩阵分解算法(LRSMD)对原始图像进行异常检测,将检测结果中的异常像元求取平均作为近似目标光谱,最后利用近似目标对原始图像进行约束能量最小化(CEM)匹配检测.为验证所提算法的有效性,分别用两幅真实高光谱图像进行仿真实验.实验结果表明,与LRSMD算法相比,新算法能够有效地抑制虚假目标,显著地提高异常目标的检测性能.
Aiming at the problem that the detection accuracy of anomaly detection algorithm is much lower than that of the target detection algorithm, a hyperspectral anomaly detection algorithm based on approximate target posterior information is proposed. The algorithm firstly uses the LRSMD (low rank and sparse matrix factorization algorithm) Anomaly detection, the abnormal pixels in the detection results are averaged as the approximate target spectrum, and finally the approximate image is constrained by the CEM (matching energy minimization) detection.For the validity of the proposed algorithm, Real hyperspectral images are simulated.The experimental results show that compared with LRSMD algorithm, the new algorithm can effectively suppress false targets and significantly improve the detection performance of abnormal targets.