Detecting moving targets in multiframe SAR imagery without using any direct backscatter

Detecting moving targets using Synthetic Aperture Radar (SAR) imagery is challenging since target images become highly distorted due to the unknown motion of the target. The distortions include smearing, and displacement, of the target signal across the SAR image. This paper reports an approach to moving target detection which avoids these distortions by operating on the target’s shadow. Unlike a target’s signal, its shadow is not subject to distortions arising from the target motion, and the shadow appears at the true target location. Our approach to target detection seeks to make target motion observable by sub-dividing the synthetic aperture, c.f. standard SAR processing. The cross range resolution of the image is therefore sacrificed so as to gain a commensurate increase in the number of SAR images. Moving shadows are then observable in the sequence of images. We perform detection and tracking jointly, and use methods that integrate both target and background information over the sequence of images. Integration over a number of frames is desirable since it allows detection at lower signal to noise than single frame methods. A new Bayesian track-before-detect scheme is developed that operates directly on the intensity levels in a residue image sequence. The residual images are computed from the measurements and comprise target shadows which stand out from the background. Robust detection and tracking has been demonstrated on real SAR imagery from two systems: a helicopter borne system, and a longer range airborne platform. Extrapolation of the empirical results indicates that shadow detection is viable at tens of kilometre stand-off range.