基于背景纯化的改进协同表示的高光谱异常目标检测

Improved Collaborative Representation for Hyperspectral Anomaly Detection Based on Backedground Refinement

摘要:

针对协同表示的高光谱异常目标检测算法的异常点敏感问题,提出了一种基于背景纯化的改进协同表示的高光谱异常目标检测算法。利用扩展数学形态学的膨胀操作消除局部背景模型中可能存在的异常点,从而得到更为纯净的背景字典,能够有效地消除检测过程中异常点对检测效果的负面影响,从而提高检测精度。采用该算法对高光谱数据进行仿真实验,并与现有算法进行对比,结果表明该算法具有更好的检测效果。

Abstract:

Aiming at the problem of the anomaly point sensitivity of collaborative representation for hyperspectral anomaly target detection algorithm, an improved collaborative representation for hyperspectral anomaly detection based on background refinement is proposed. The expansion of mathematical morphology is used to eliminate the anomaly points that may exist in the local backedground model, so that a more pure backedground dictionary can be obtained, which can effectively eliminate the negative influence of anomaly points on the detection and improve the detection accuracy. The algorithm is applied to make simulations on hyperspectral data and compared with the existing algorithms, and the results show that it can realize a better detection effect.