Layered Target Reacquisition and Prosecution from Wide Area Motion Imagery (WAMI) Cues

ABSTRACT: This Small Business Innovation Research Phase I project develops innovative methods for target handoff from WAMI to SUAV sensors in the presence of geo-registration errors and target ambiguities. The tiered computational framework provides increasing accuracy as a function of available computational power and communication bandwidth. The lowest tier with minimum complexity uses target geo-location, kinematics and features such as shape and size for target handoff. The second tier incorporates target appearance with geometric and environmental invariant/covariant target models. In the absence of truly invariant features or models, it selects the optimal features that are discriminative in the given scenario. The next tier exploits target context in terms of neighboring targets and scene information to resolve ambiguities. The highest tier uses a feedback mechanism from the SUAV to improve handoff certainty and to recover from association errors. The project further offers analysis of target handoff algorithms with respect to a wide variety of factors including sensor properties, communication parameters, environmental artifacts, available computational power, and scene elements. The Phase I effort will include: development of enabling algorithms, implementation of the framework, demonstration of proof of concept, and parametric and quantitative evaluation of the proposed technologies using real and synthetic WAMI/SUAV video. BENEFIT: Wide-area motion imagery has proven to be of tremendous use in exploitation and target tracking on ground as it offers coverage of large areas for long durations. Smaller UAVs, on the other hand, provide much better resolutions on targets but with limited field-of-views. Target handoff from wide-area motion imagery (WAMI) sensor to sensors on small unmanned aircraft systems (SUAS) or weapons can provide autonomous collaborative and persistent surveillance and reconnaissance using networked unmanned platforms. However, it is a challenging problem due to variations of sensing geometries, sensor modalities, and imaging parameters, the uncertainties in sensor metadata and geo-localization of target in both WAMI and SUAS imagery, varying object density and dynamics, and environmental factors such as obscuration. The tiered approach presented in this proposal aims to overcome these challenges with the following benefits: High resolution coverage of high-value targets in the scene. Precision targeting in cluttered areas in the presence of accompanying targets. Maximization of information content for situational awareness. Collaborative sensing for timely and consolidated situational awareness and visualization. Resource management by identifying the most suitable SUAV for tasking from a pool of available sensors.