Textural feature based target detection in through-the-wall radar imagery

Stationary target detection in through-the-wall radar imaging (TWRI) using image segmentation techniques has recently been considered in the literature. Specifically, histogram thresholding methods have been used to aid in removing the clutter, resulting in ‘clean’ radar images with target regions only. In this paper, we show that histogram thresholding schemes are effective only against clutter regions, which are distinct from target regions. Target detection using these methods becomes challenging, if not impossible, in the presence of multipath ghosts and clutter that closely mimics the target in size and intensity. Because of the small variations between the target regions and such clutter and multipath ghosts, we propose a textural feature based classifier for through-the-wall target detection. The feature based scheme is applied as a follow-on step after application of histogram thresholding techniques. The training set consists of feature vectors based on gray level co-occurrence matrices corresponding to the target and ghost/clutter image regions. Feature vectors are then used in training a minimum distance classifier based on Mahalanobis distance metric. Performance of the proposed scheme is evaluated using real-data collected with Defence Research and Development Canada’s vehicle-borne TWRI system. The results show that the proposed textural feature based method yields much improved results compared to histogram thresholding based segmentation methods for the considered cases.