Tracking and detection of moving point targets in noise image sequences by local maximum likelihood

A pair of tracking and detection algorithms are proposed. The tracking is accomplished by many tracking blocks that have a size equal to the maximum velocity of the moving target. At the initial stage, the tracking blocks cover the whole image plane without overlap. A a tracking block adaptively moves so that its center moves to the pixel that has the largest intensity in the region covered by the block and the largest intensity is recorded. The detection is based on the pixel intensities recorded by each tracking block. The asymptotic probability that any pixel in the image plane is covered by some tracking blocks is shown to be l-e/sup -1/. The performance is analyzed and is fairly high. The results of simulation tests confirm to the theoretical analysis.