Kernel oblique subspace projection approach for target detection in hyperspectral imagery

Abstract

In this paper, a kernel-based nonlinear version of the oblique subspace projection (OBSP) operator is defined in terms of kernel functions. Input data are implicitly mapped into a high-dimensional kernel feature space by a nonlinear mapping, which is associated with a kernel function. The OBSP expression is then derived in the feature space, which is kernelized in terms of the kernel functions in order to avoid explicit computation in the high-dimensional feature space. The resulting kernelized OBSP algorithm is equivalent to a nonlinear OBSP in the original input space. Experimental results based on simulated hyperspectral data and real hyperspectral imagery shows that the kernel oblique subspace projection (KOBSP) outperforms the conventional OBSP. © 2010 Springer-Verlag.