A Robust Real-time Moving Object Tracking Algorithm

The fast and reliable way to measure the similarity of tracked targets in sequence frames is the key point in visual tracking. Histogram is one of stable character of objects, which does not vary with change of target shape and scale. As the histogram does not contain the spatial information for the target, there comes out many pseudo-targets with similar histogram with the matched object. Besides the histogram describes color distribution in one region. It results in that the nearby regions have the similar color distribution, which make it difficult to locate target precisely. There always exists some distance away from the best position. On the other hand, the template matching is other useful way to measure the targets, but it is affected too much when target shape changing or overlapped. In this paper we propose a robust object tracking algorithm based on both histogram and template character. The histogram is used as object basic feature, and the flexible template is used for validating the reliability of candidate objects. And Mean shift algorithm is used in the histogram matching to speed up the algorithm to meet the real time requirement. Finally, the paper gives the experiment results, which show that the algorithm is practical and efficient.