视频监控中运动物体跟踪算法研究

Research on Moving Object Tracking Algorithm in Video Surveillance

笔者提出基于双模板的视频监控中运动物体跟踪算法,该方法组合使用Meanshift与Kalman滤波,并增加双跟踪模板进行跟踪矫正。对视频中的运动物体进行MeanShift跟踪,将跟踪的结果与设定的双跟踪模板进行相似性度量。为减少跟踪的偏移误差,在与双模板进行相似性度量前,判断长跟踪模板是否到达更新周期。若跟踪目标与双模板度量的结果在设定的阈值内,则用当前跟踪结果作为新的跟踪窗口;若度量的结果不在设定的阈值范围内,则启用Kalman滤波跟踪法;而当Kalman滤波跟踪得到的结果也不在设定的阈值内时,放弃跟踪序列。实验结果表明,该算法在现实视频场景下跟踪的稳定性较强,有效减少了漂移现象。

This paper proposes a dual-template-based tracking algorithm for moving objects in video surveillance.This method combines Meanshift and Kalman filtering,and adds dual-tracking templates for tracking correction.Perform MeanShift tracking on the moving objects in the video,and measure the similarity between the tracking result and the set dual tracking template.In order to reduce the tracking offset error,before the similarity measurement with the dual template,judge whether the long tracking template has arrived update cycle;if the tracking target and the results of the dual template measurement are within the set threshold,the current tracking result is used as the new tracking window;if the measurement result is not within the set threshold,the Kalman filter tracking method is enabled,and when the Kalman When the result of filtering tracking is not within the set threshold,the tracking sequence is abandoned.The experimental results show that the algorithm has strong tracking stability in real video scenes and effectively reduces the drift phenomenon.