博物馆监控视频中慢速移动稀疏目标异常轨迹检测研究

传统异常轨迹检测方法将轨迹序列看作轨迹特征,无法有效描述轨迹,导致异常轨迹检测结果不可靠。为此,提出一种新的博物馆监控视频中慢速移动稀疏目标轨迹检测方法。采用一种快速计算方法对和目标相似度较高的粒子进行筛选,滤除和均值相差较大的粒子。对跟踪目标进行稀疏表示,为了避免目标被干扰或遮挡,进行非负性约束优化,完成稀疏求解,获取博物馆监控视频中慢速移动稀疏目标跟踪结果。依据跟踪结果将可代表整体轨迹的特征向量与部分可代表局部轨迹的特征向量合成一个整体特征向量,利用整体特征向量对慢速移动稀疏目标轨迹进行描述,通过描述结果和K聚类方法实现目标异常轨迹检测。实验结果表明,所提方法检测的异常轨迹与其他轨迹之间的差异最大,检测结果可靠,实际应用性较高。

Slow moving sparse target anomaly detection in Museum monitoring video

The traditional anomaly trajectory detection method takes the trajectory sequence as the trajectory feature, and can not effectively describe the trajectory, which leads to the unreliable result of abnormal trajectory detection. To this end, a new slow moving sparse target trajectory detection method in Museum monitoring video is proposed. A fast calculation method is used to filter the particles with higher similarity to the target and filter out the particles with larger difference in the mean value. Sparse target tracking, in order to avoid interference or occlusion of target and non negative constrained optimization, complete sparse solution, slow moving target tracking results to obtain sparse feature vectors in video surveillance Museum; synthesis of a whole feature vector and part of the basis for tracking results will represent the overall trajectory can represent the local path and describe the slow moving target trajectory using whole sparse feature vector, to achieve the goal of trajectory anomaly detection by describing the results and K clustering method. The experimental results show that the difference between the anomaly trajectory and the other trajectories detected by the proposed method is the greatest, and the detection results are reliable and the practical application is high.