IMSAR Near-Real-Time, Onboard ATR for Low-SWaP Platforms

Radar provides valuable, time-sensitive Intelligence, Surveillance, and Reconnaissance (ISR), however, the identification of targets of interest can be difficult in radar data. Although analysts can be trained to provide target recognition from imagery, the work is time-consuming and prone to human error. Contrarily, Automatic Target Recognition (ATR) algorithms can be trained to accurately identify and classify objects in Synthetic Aperture Radar (SAR) radar imagery. Because traditional ATR systems are large platforms, they are unsuitable for the stringent size constraints of typical ISR Unmanned Aerial Vehicles (UAV) operations. To provide a light-weight, real-time ATR option, IMSAR proposes to adapt its commercial low-Size, Weight, and Power (SWaP) radar systems to automatically identify targets. Specifically, IMSAR proposes to port existing, model-based ATR code to a code-base that can run on a small form-factor GPU processor onboard the radar system. IMSAR will partner with a subcontractor to optimize their model-based ATR software with radar data. On-board ATR will provide real time processing, allowing ATR to run on a CUDA-enabled processor alongside existing software on the airborne radar. Coupled with IMSAR’s family of NanoSAR radars and open API, these advance ATR algorithms will provide platform agnostic exploitation.