Georegistration of Imagery with Target Tracking

Precise detection, tracking and geo-location of multiple moving and stationary targets from airborne sensors in real time poses a great challenge due to the separate registration tasks involved, with the most difficult being geo-registration which requires matching images taken under different conditions. Registration of one image to the next in a video sequence allows detection and tracking of moving targets by aligning common background features in the images, but to obtain the geo-location of these targets the image must also be registered to a geodetically calibrated reference image and overcome the potentially drastic differences in viewing angle, elevation and season to align common world feature points. Toyon proposes to design algorithms to unify frame- and geo- registration by combining the feature point analysis necessary for image registration. Toyons approach to computationally efficient geo-registration is based on a detailed understanding of automated video analysis, and will minimize complexity by only updating a frame to the geo-reference image when the scene content has significantly changed, mapping the known pixel geo-coordinates between intermediate frames. When fully developed in commercially available hardware, this technology is expected to provide significant improvements in exploitation of aerial video surveillance data, in real time. BENEFIT: Successful completion of the proposed research and development effort will result in a real-time video-based target detection and tracking system with track outputs in world coordinates and geo-referenced output frames.  The main benefits are expected in military applications where coordination among multiple airborne and ground assets is necessary and real-time performance is critical.  However, smaller scale operations would be able to use these capabilities along borders, near ports, or along the coastline to monitor for illegal or terrorist activities. The proposed techniques can be extended to land vehicle-born implementations for enhanced situational awareness along highways or country roads.  Potential commercial applications include many police or homeland security applications.