A nonparametric F-distribution anomaly detector for hyperspectral imagery

An innovative idea is proposed and its application to hyperspectral imagery is presented, as a viable alternative to testing sample hypothesis using conventional methods. This idea led to the design of two novel algorithms for anomaly detection. The first existing algorithm, referred to as semiparametric (SemiP), is based on some of the advances made on semiparametric inference. The second algorithm, proposed in this paper and referred to as a combined F test (CFT), is based on a nonparametric model and has its test statistic behaving asymptotically under the Fisher's F family of distributions. A major drawback of the SemiP detector is its dependence on a function maximization routine, which requires initialization and no guarantees of convergence. The CFT detector is free of such dependence. Experimental results using real hyperspectral data are presented to illustrate the effectiveness of both algorithms in comparison to the industry standard approach. The CFT and SemiP detectors significantly outperformed the standard approach