EO/IR Multi-Sensor Fusion Tracker Algorithm

Toyon Research Corporation proposes research and development to maximize exploitation of co-boresighted multi-sensor EO/IR video, based on an innovative Bayesian track-before-detect algorithm. Statistical background modeling is performed to optimize clutter suppression, and estimated moving target likelihoods are integrated over time, without thresholding, in a non-linear particle filter. Toyon"s proposed algorithm includes multiple-target feature-aided data association via a novel Multiple-Frame Soft-Assignment Solver (MFSAS) developed in a Belief Propagation framework for computationally efficient estimation of probabilistic associations in large-scale tracking problems. Following successful development in Phase I and II, the resulting EO/IR multi-sensor fusion and tracking device is expected to be capable of real-time airborne processing of multiple unregistered video feeds within EO/IR turret SWaP constraints, based on a parallel processing architecture composed of a heterogeneous mix of embedded processors.