Integration of Bayes detection and target tracking in real clutter image sequences

We present an optimal Bayesian algorithm for integrated, multiframe detection and tracking of dim targets that move randomly in spatially correlated, cluttered image sequences. The algorithm consists of a multiframe minimum probability of error Bayes detector integrated with a multiframe maximum a posteriori (MAP) position estimator. The design of the detector/tracker incorporates the models for target signature, target motion, and clutter; it uses recursive spatio-temporal processing across all available frames to make detection decisions and to generate position estimates. A simulation with an artificial target template added to a real clutter background shows that the proposed algorithm outperforms the association of a standard single frame image correlator and a linearized Kalman-Bucy filter in a scenario of heavy clutter.