Image-based Tracking and Sensor Resource Allocation for Unmanned Aerial Vehicles

A key objective of our project is to develop advanced algorithms and software to track people and vehicles in an urban environment using video imagery from multiple unmanned aerial vehicles (UAVs). Track continuity of targets of interest (TOI) is an essential requirement. Collaborative UAV trajectory planning and camera pointing to collect continuous observations on one or more TOI by a team of UAVs are required when the line-of-sight of a video camera is occluded by buildings or trees. In the Phase I, we developed advanced feature-aided video tracking algorithms and software using the multiple hypotheses tracking (MHT) approach. We used real video data and developed algorithms and software for realistic video measurement model, geolocation, track initiation, and nonlinear filtering. In Phase I, we processed real UAV video data and obtained preliminary results. In Phase II, we shall process a large number of datasets with single and multiple UAVs. We propose to develop sensor resource management algorithms and software by leveraging on an AFOSR funded project on dynamic UAV trajectory planning and camera pointing to mitigate the occlusion of the camera line-of-sight.