多路视频智能场面监视系统的关联决策算法及应用

The application of association decision algorithm in multi-sensor video surveillance and monitor system

多路视频智能场面监视系统中目标运动机动性强、方向和速度变化大、轨迹交叉和分岔多,并且目标身份通常无法确定,因而无法通过二者的目标身份信息进行确切关联,关联决策具有极大的不确定性。本文在计算和分析两条轨迹之间的4种距离及其统计特征的基础上,根据给定的漏关联概率和错误关联概率确定每种距离对应的决策门限,最后进行加权平均计算出综合距离和综合门限,由此构造关联决策规则。在给定的漏关联概率和错误关联概率条件下,对两个视频传感器的轨迹数据运用关联决策规则,试验结果证明,能够满足场面监视系统的基本要求。

In multi-sensor video surveillance and monitor (VSAM) system, objects have strong maneuverability such as fast change of speed and direction, so their trajectories are often cross and fork, and identify of objects can't be recognized also, so trajectories can't be associated exactly. According the attributes data of trajectory, and based on calculating four type distances and respectively thresholds and weights between two trajectories, an associate decision algorithm is presented here to construct the associate decision rules. According the special probability of loss association and the special probability of virtual association, the results of 2-sensor VSAM are obtained and show the algorithm can be used in VSAM.