投影追踪方法在高光谱图像异常点检测中的应用

Application of Project Pursuit in Hyperspectral Anomaly Detection

该文提出了一种基于投影追踪的高光谱图像异常点检测方法.它通过广义似然比检验(GLRT)模型构建二元检测算子,并利用观测数据估计出算子中代表背景的未知参数,而算子的关键参数——目标参数是通过投影追踪算法搜索异常点得到的.此算法消除了传统的基于多元统计模型的目标检测方法对先验信息的依赖。增强了算法的实用性.同时,投影追踪方法能有效的提取目标参数,进一步提高了异常点检测的效果.

In this paper, a new method of hyperspectral anomaly detection based on project pursuit is presented. The Generalized Likelihood Ratio Test(GLRT) is used to establish a binary hypotheses detector and estimates the unknown parameters that represent the background in the detector from the image. Target information, the key parameter, is got by using project pursuit approach to search anomaly information. The algorithm reduces the dependence of pre-information, enhances the arithmetic practicability. At the same time, project pursuit approach can extract target information efficiently and improve the effect of anomaly detection.