基于区域模板匹配的多目标实时跟踪

摘要

In intelligent video surveillance systems, moving targets tracking is the foundation of follow-up work.In order to track multi-targets in real-time and accurately, a rapid multi-targets tracking method combined with improved inter-frame difference and region template matching was presented in this paper. Firstly, areas of moving targets are detected according to the improved inter-frame difference, then, searching areas are constructed. Secondly,if the searching area is matched with some pre-existing target template, mark the corresponding target with an ID number, and then update the template. Then the multi-targets are tracked. In the experiment, the average False Positive Rate of three targets is 11.11％, the result shows that the method is effective and can basically meet the requirement of real-time target tracking.%研究视频目标实时跟踪问题,在智能视频监测系统中,对运动目标的跟踪是后续工作的基础.为了实时准确地跟踪多运动目标,提出一种基于改进帧间差分与区域模板匹配相结合的快速运动目标跟踪方法.根据改进的帧间差分法找到运动目标的所在的区域,建立相应的搜索区域,采用目标模板逐个与搜索区域进行匹配,匹配到目标后对其进行标记,并且对相应的目标模板进行更新,从而实现多个目标的跟踪.对三个目标进行跟踪平均的第Ι类错误率为11.11%,结果表明,改进方法能够比较有效、实时地对多个运动目标进行跟踪.