Deep Learning Stationary Target Detector and Classifier

To address the Army need for an algorithm for enhanced detection and classification of stationary ground targets for fire control radar (FCR) mounted on Apache attack helicopters, Physical Optics Corporation (POC) proposes to develop a new Deep Learning Stationary Target Detector and Classifier (DEESTAC) technology. DEESTAC is based on a customization of recent advancements in deep-learning methods for classifying objects in images. Specifically, the innovation in deep-learning convolutional neural networks (CNNs) to enhance radar resolution and detect stationary targets with complex radar signatures will enable DEESTAC-based radar systems to reliably detect and classify targets in data with high signal-to-noise ratios (SNRs). Thus, this technology offers improved battlefield awareness for an appropriate warfighter response in life threatening situations, which directly addresses the Army need. In Phase I, POC will demonstrate the feasibility of DEESTAC by developing a technology readiness level (TRL)-3 prototype. We will investigate the feasibility of DEESTAC by developing its algorithms and testing them on radar data. In Phase II, POC plans to build and deliver a TRL-6 prototype, implemented and tested with extensive radar data representing a variety of target scenarios.