Hyperspectral Target Detection Using Kernel Matched Subspace Detector (WP-P5)

Author(s) : Heesung Kwon (US Army Research Laboratory, USA) Nasser M. Nasrabadi (US Army Research Laboratory, USA)

Abstract : In this paper we present a nonlinear realization of a signal detection approach that uses the generalized likelihood ratio tests (GLRTs). It is based on converting the linear subspace models, so called matched subspace detectors (MSD). The linear models for the GLRTs of MSD are first extended to a high dimensional feature space and then the corresponding nonlinear GLRT expressions are obtained. In order to address the intractability of the GLRTs in the nonlinear feature space we kernelize the nonlinear GLRTs using kernel eigenvector representations as well as the kernel trick where dot products in a nonlinear feature space are implicitly computed by kernels. The proposed kernel-based nonlinear detectors, so called kernel matched subspace detectors (KMSD), are applied to a given hyperspectral imagery -- HYDICE images -- to detect targets of interest. KMSD showed superior detection performance over MSD for the HYDICE images tested in this paper.

Menu