Random-Selection-Based Anomaly Detector for Hyperspectral Imagery
Random-Selection-Based Anomaly Detector for Hyperspectral Imagery
复制标题
基于随机选择的高光谱图像异常检测器
DOI：
10.1109/tgrs.2010.2081677
复制
发表时间：
2011-05-01
影响因子：
8.2
通讯作者：
Zhang, Liangpei
中科院分区：
文献类型：
--
作者：
Du, Bo;Zhang, Liangpei
高光谱图像中的异常检测是目标检测领域的一个重要研究方向，因为它不需要先验信息，充分利用了高光谱图像中的光谱差异。目前的异常检测方法容易受到处理窗口范围或图像范围内的异常的影响。此外，对于局部异常检测方法本身，很难确定适合处理背景统计的窗口大小。提出了一种基于随机选择背景像素的异常检测方法--基于随机选择的异常检测器（RSAD）。从图像场景中随机选择像素来表示背景统计;随机选择被执行足够的次数;在选择背景像素的适当子集之后，每次使用阻塞的自适应计算高效的离群点提名器来检测异常;最后，采用融合过程来避免异常像素对背景统计的污染。此外，通过对更新数据的随机选取和QR分解，实现了RSAD的实时实现。在实验中使用了多个高光谱数据集，RSAD表现出比现有的高光谱异常检测算法更好的性能。实时版本的性能也优于实时版本。
Anomaly detection in hyperspectral images is of great interest in the target detection domain since it requires no prior information and makes full use of the spectral differences revealed in hyperspectral images. The current anomaly detection methods are susceptible to anomalies in the processing window range or the image scope. In addition, for the local anomaly detection methods themselves, it is difficult to determine the window size suitable for processing background statistics. This paper proposes an anomaly detection method based on the random selection of background pixels, the random-selection-based anomaly detector (RSAD). Pixels are randomly selected from the image scene to represent the background statistics; the random selections are performed a sufficient number of times; blocked adaptive computationally efficient outlier nominators are used to detect anomalies each time after a proper subset of background pixels is selected; finally, a fusion procedure is employed to avoid contamination of the background statistics by anomaly pixels. In addition, the real-time implementation of the RSAD is also developed by random selection from updating data and QR decomposition. Several hyperspectral data sets are used in the experiments, and the RSAD shows a better performance than the current hyperspectral anomaly detection algorithms. The real-time version also outperforms its real-time counterparts.