Study Of The Video Object Tracking Algorithm Based On Weighted Color Histogram

In computer vision, image processing and pattern recognition, tracking thedifferent objects after detecting is an important content and a large field in continuousvideo sequences. The system based on object tracking in continuous video sequencesis extensive, such as monitoring system, transportation system, guidance system,navigation system. Therefore, Target detection and tracking in continuous videosequence has a lot of practical and applied value. This paper studies the problem ofsingle target tracking in continuous video sequences, the quality situation of thebackground weighted MeanShift algorithm in target tracking is discussed, andimprove the algorithm, at the same time it extend to multiple target tracking afterdetecting, the problem of tracking multiple targets is divided into five parts，whichmakes a concrete analysis. The application of the background is the monitoringsystem. The main work of this paper is as follows:The first, this paper studies the background weighted MeanShift algorithm basedon the original MeanShift algorithm, the algorithm of background weightedMeanShift is proposed in order to reduce the tracking target background informationin the MeanShift tracking algorithm. However, the background weighted MeanShiftalgorithm reduced the background information around the target, but in the MeanShiftiteration convergence calculation formula of position has not changed, this paper putforward respectively execute in the initial target region and the target candidateregion, in the calculation of the initial regional execute background weightedalgorithm, and execute an adaptive block search algorithm in the target candidateareas, the algorithm of this paper is faster on convergence speed, more accuratelocation of calculation, stronger robustness than the traditional MeanShift algorithmand background weighted MeanShift algorithm.The second, according to the complex object tracking environment, the shelterbetween targets or targets and background occlusion, the random state of trackingtargets. For such a perplexing situation, the target tracking problem is divided into five aspects: the new target, normal target tracking, target missing, multiple targetfusion, a single target separation, and this paper match the separation object throughKalman filter and the amendment color matching matrix in this complex situation, inorder to improve the accuracy of the system, ensure continuous tracking, enhance therobustness of the system.Finally, through the analysis of the experiment, it proves that the improvedalgorithm has a faster convergence speed, and in multiple tracking targets,recognizing the target after splitting is better, and it has strong robustness in theprocess of target tracking.