Multitarget Tracking By Distributed Cooperative Processing ( Extended

Exploiting a new distributed cooperative processing scheme where multiple processors cooperate in finding a global minimum, we have developed a new efficient maximum likelihood-based calculation method for multitarget motion analysis under a fixed networked multisensor environment. The marked improvement in computational efficiency and also in stability is achieved by replacing the well known Hungarian assignment algorithm [13, 11] with a much simpler sorting algorithm of O(N logN) and fusing the result with locally minimized average square errors. We have proved a theorem which asserts that an optimal data assignment matrices can best be given in terms of sorted bearing measurement data of targets. Embedding an optimal data association algorithm of O(N logN) into each of GausNewtons’s downhill iteration loops, our numerical experiments were able to track as many as 8 targets and 12 targets separately within one minute by 400MHZ Dell computer with improved accuracy and efficiency, where all targets are allowed to move in variable directions at varying speeds if 4 and 6 processors are used respectively. keywords Cooperative distributed vision, data associate problem, multitarget motion tracking