A Loop-Template Matching Algorithm for Target Tracking Based on Kalman Filter

The global searching method is adopted in the traditional correlation algorithms in target tracking system, resulting in the large amounts of calculation, the difficulty of real-time implementation and even the loss of the tracking target when the target is partially shaded or in sharp attitude variation. To solve the problems, an improved approach based on the loop correlation template matching and Kalman filter is proposed to realize target tracking. Kalman filter, to make the tracking more active, estimates the possible region of the target in the next frame at first,and finds the best matching points by the loop correlation matching in the aforementioned region while the loop-template matching overcomes the loss of the horizontal matching points caused by the sharp attitude variation. So the algorithm has strong robustness, and is of translation and rotation invariance. In the experiments, the proposed algorithm is applied to the tracking of a main battle tank partially shaded and a jet with "Viktor Pugachev's Cobra" . Compared with the traditional algorithms, the algorithm herein can steadily track the targets in less time without losing them.