Real-Time Exploitation of Video Synthetic Aperture Radar Imagery

Aret Associates proposes the video-SAR tracker (VST) that combines 3 existing Aret capabilities in response to the stated need to detect, track and classify all moving targets with particular emphasis on slow-walking humans using video-SAR. The VST employs 1.) the existing Aret Radar Image Processor (ARIP) to process video-SAR data and to apply noncoherent change detection techniques to identify SAR chips of potential movers; 2.) the deep learning engine that Aret has demonstrated to have detection and classification performance that significantly outperforms template-based filter approaches; and 3.) a particle filter tracker to integrate the detection, tracking and classification of the movers. The core deep learning application contains a framework that permits integration of physical understanding, prior information and will generate significant false alarm reduction. The proposed effort will produce a feasibility demonstration with existing, available SAR data and contains a clear transition path towards implementation with the APS-153.