Low-shot Object Tracking and Targeting Algorithms for Next-generation Combat Vehicles

Physical Sciences, Inc. (PSI) proposes to develop the Low-Shot Object Tracking and Targeting (LOTT) system, which integrates state-of-the-art machine-learning object detection and tracking capabilities with thermal IR and visible imaging sensors utilizing “low-shot” training techniques. LOTT algorithms will enable the detection, characterization, and tracking of an easily-updated target set of objects in the complete 360 degree environment around modern, instrumented combat vehicles. These algorithms will be optimized for and demonstrated on embedded processing platforms. A complete design including required processing hardware, algorithms, interface definitions, and compatible sensors will be developed. The LOTT algorithms will fuse all available, stationary and slewable, single-channel and multi-spectral thermal and visible EO imaging systems into a complete situational awareness view. This capability will leverage an existing, PSI-developed object detection and tracking framework for visible EO cameras that executes on low SWaP embedded platforms, but expand this framework to operate in the thermal IR. The system will assist soldiers in detecting and prioritizing threats and provide reliable tracking output to enable slewable sensors to maintain targeting on selected objects. The proposed development will maximize the effectiveness of deployed and planned sensor systems while minimizing associated operator burden.