A target alignment algorithm for through-the-wall radar imagery classification

Sensing through the wall using radar is a valuable capability. There is considerable work in generating radar images of the interior of a room by beamforming of the radar backscatter generated in a Synthetic Aperture Radar (SAR) configuration. However, high level interpretation of the scene is a more difficult task. In previous work a minimum distance classifier was successfully used to recognize various targets placed in the scene. The approach suffered from the dependency of target features on target location. This work presents a solution to this problem by bringing the target intensity profiles into alignment with the training data prior to classification. The alignment is performed by moving the intensity profiles between locations using an Autoregressive Moving Average (ARMA) model. The classification results before and after alignment show a marked improvement.