Local covariance matrices for improved target detection performance

Abstract

Our research goals in hyperspectral point target detection have been to develop a methodology for algorithm comparison and to advance point target detection algorithms through the fundamental understanding of spatial/spectral statistics. In this paper, we demonstrate improved target detection performance by making better estimates of the covariance matrix. We develop a new type of local covariance matrix which can be implemented in Principal Component space which shows improved performance based on our metrics.