Global, Local, and Stochastic Background Modeling for TargetDetection in Mixed Pixels

摘要

As hyperspectral sensors and exploitation methods have evolved, the accuracy of conventional background models hasbecome a limiting factor for high confidence and low false alarm detection of mixed pixel targets. Many common targetdetection algorithms, such as the Adaptive Coherence/Cosine Estimator, implicitly use a global background model thatassumes the background can be modeled by a single, multivariate Gaussian random variable with additive independentand identically distributed Gaussian noise. In order to improve the accuracy of the Gaussianity assumptions, a localbackground model, which models the background as multiple, disjoint Gaussian clusters, has also been widelyconsidered. This paper introduces an improved variant of the local background model, as well as a novel stochasticbackground model that is free from distributional assumptions and accounts for the spectral variability of the backgroundon a pixel-by-pixel basis. The performance of the global, local, and stochastic background models is evaluated on acontrolled data set and the tradeoffs associated with each method are discussed.