Persistent Electro-Optical/Infra-Red (EO/IR) Wide-Area Sensor Exploitation

Toyon proposes to develop high-performance algorithms and real-time software for WAPS video exploitation in accordance with AFRL’s Voxelpedia concept of long-term statistical and syntactic learning in a geo-registered 3D framework. Toyon’s approach is based on dense 3D reconstruction of every pixel in the scene using a novel algorithm which provides significantly improved modeling accuracy, while maintaining computational efficiency required for airborne processing. Toyon’s approach to moving target detection is based on statistical background modeling for suppression of motion clutter, and building the model in the 3D framework enables accurate registration to new video frames in the presence of parallax effects, effectively mitigating the greatest challenge to automated moving target detection in urban environments. Numerous other improvements to WAPS video exploitation are also enabled, including persistent target tracking and identification based on super-resolved features, statistical characterization of normal target motion and identification of anomalous motion, and rule-based alerting of possible threatening target actions defined and observed in the context of the geo-registered 3D framework. In Phase II, Toyon proposes to perform large-scale statistical characterization of the software performance, document and deliver the developed algorithms and software to AFRL, and support integration of the developed technology in AFRL’s TechEdge laboratory. BENEFIT: The successful completion of this research and development will result in significant improvements in real-time exploitation of persistent wide-area high-resolution video-rate EO/IR imaging sensors, including automated target detection, tracking, and identification in dense and cluttered urban environments. Secondary technical benefits will include the ability to automatically calibrate imaging sensors using flight data and rapidly create high-fidelity geo-registered 3D models of large scenes. This technology would have direct applications in deployed and current developmental wide-area persistent surveillance EO/IR sensor systems, including Angel Fire, Constant Hawk, ARGUS-IS, Gorgon Stare, and TAILWIND, as well as in a broad range of Air Force and other DoD surveillance systems, including air-, ground-, sea-, and space-based platforms. In surveillance applications, geo-registered 3D data provides context for the sensor operator, as well as exploitation system operators. Highly precise 3D geo-registration is required for multi-look change detection to locate emplacement of possible structures of interest or of threats such as IEDs, and on shorter time scales to perform moving-target detection. The technology could also be used to provide improved realism in personnel training, and even in computer gaming.