Automatic target recognition organized via jump-diffusion algorithms

Random sampling algorithm based on jump-diffusion dynamics enables simultaneous detection, tracking and automatic target recognition in air-to-ground and ground-to-air scenarios. A Bayesian approach is adopted for inference of scenes. New objects are detected, and discrete jump moves through parameter space are studied for identification of objects. The algorithm explores scenes of different complexities. Continuous diffusions generate scale and rotation group transformations between jumps. The method combines dynamics-based tracking within one estimation framework.