Processing of Large Wide Area Airborne Sensor (WAAS) Data Streams in Hardware

ObjectVideo and Brown University propose a heterogeneous FPGA and programmable multi-core DSP solution to perform full-frame target tracking and event detection on motion imagery from Wide Area Aerial Sensors. The OV team will leverage its experience in WAAS video processing and low-power hardware design in order to provide excellent tracking and event detection performance while meeting the size, weight, and power goals appropriate for inclusion on small unmanned aircraft systems. The team has developed local feature-based target detection and tracking algorithms that operate on a significantly reduced space of interest points. These exploitation algorithms scale to larger frame sizes better than conventional techniques and have proven effective on aerial imagery. The team will evaluate these algorithms for execution on FPGA and DSP hardware, and formulate a scalable model for power, area, performance and precision which can guide the hardware implementation. This solution utilizes technologies and methodologies the OV team developed under DARPA funding to support the ARGUS-IS and IR projects; these technologies can be adapted for this effort with minimal risk.