Two-stage process for improving the performance of hyperspectral target detection

Abstract

The spectrum of each pixel in a hyperspectral image usually comprises multiple material spectra, due to the sensor's spatial resolution and ground material distribution. The purpose of target detection (TD) is to separate specific target pixels from the various background pixels, using a known target signature. In this paper, a novel two-stage target detection process is proposed for improving TD performance. In the first stage, a target detector is applied. In the second stage, the detected result is sorted in ascending order, a portion of the ascending data is selected, and the target detector is reapplied using the selected subset data. In this study, three real hyperspectral data-cubes with ground truth and two well-known target detectors are used to evaluate and compare the performance of this method. The experimental results show that the proposed two-stage TD process improves the detection quality, with a reduced number of false alarms.