基于目标跟踪的群聚行为识别

摘要

For video surveillance in the flocking behavior, proposed clustering behavior recognition method based on target tracking. When the number of targets in the scene is not much, and no severe blocking each other, through a combination of frame difference, CamShift color tracking algorithm, template matching and Kalman predicted to form a multi-sensor fusion idea of target tracking algorithm can achieve movement target tracking and multi-threaded implementation of the multi-target tracking. In addition, the fitting can also track each moving target and the predicted trajectory of the next time period, and the trajectory properties of the statistical distribution of all targets in the image area of the movement trend ultimately by observing whether a region belonging to the target time possible gathering area, in order to determine whether the event occurred cluster. Experimental results show that this method is suitable for the small number of scenes in the scene, and the clustering to identify better.%针对视频监控中的群聚行为，提出一种基于目标跟踪的群聚行为识别方法。当场景中目标数目不多，且相互之间没有严重遮挡时，通过结合帧差法、CamShift颜色跟踪算法、模板匹配法与Kalman预测，形成了多传感融合思路的目标跟踪算法，可实现运动目标的跟踪，并采用多线程实现对多目标的跟踪。此外，还可拟合每个运动目标的轨迹和预测下一时间段的轨迹，并统计所有目标的运动轨迹在图像区域中的运动趋势的分布特性，最终通过观察某个区域是否长时间属于目标可能的聚集区，从而判断群聚事件是否发生。实验结果表明，该方法适用于场景中人数不多的情景，且群聚识别效果较好。