The Optimal 2D Multiframe Detector/Tracker

We detail in this paper the implementation of the optimal Bayes multiframe detector/tracker for rigid objects moving randomly in two-dimensional (2D) finite grids. We present 2D models for target signature and target motion that build an integrated framework for detection and tracking. We model the background clutter by 2D correlated noncausal Gauss-Markov fields of arbitrary order. By exploring the structure of the signature, motion, and clutter models, we indicate how substantial computational savings can be achieved in the implementation of the algorithm. The detection performance of the proposed Bayes scheme is evaluated through Monte Carlo simulations. The results show significant performance gains of over 6 dB in peak signal-to-noise ratio when the optimal multiframe detector is compared to the optimal single frame likelihood ratio test (LRT) detector.