一种用于多摄像机无重叠视域环境下的车辆跟踪系统

针对当前交通监控系统中自动化程度较低、过多依赖人力进行监控的现状,文章设计了一种用于多摄像机无重叠视域环境下的车辆跟踪系统.该系统为中心-分布式结构,中心服务器与分布在前端的智能摄像机节点通过GPRS相连,与传统的功能单一的摄像头相比,基于DM642的智能节点可以在前端完成目标检测、跟踪以及分类等工作,并将抽象化后的目标特征信息通过GPRS传输给中心服务器,大大减轻了网络传输的压力.中心服务器根据节点传送来的目标信息,对不同摄像机视域下的目标进行匹配,实现了目标的连续跟踪,获得了更为全面的目标信息.试验表明,文章设计的车辆跟踪系统可以在较大监控范围内有效地跟踪目标车辆.

A Vehicle Tracking System Based on Multiple Cameras across Non-overlapping Views

A vehicle tracking system based on multiple cameras across non-overlapping views is designed aiming to solve some problems existing in traffic surveillance, such as low automation degree and relying too much on human resource. The proposed system is comprised of intelligent nodes and a central server, which are connected using GPRS. Compared with the traditional camera used in the front, the intelligent node can fulfill tasks such as target detection, target classification and target tracking based on DM642. The highly summarized information of the target is transmitted to the central server via wireless network with low transmission burden. The central server match targets in disjoint views according to the target＇s information. The camera topology is learned using the matching information and then the topology can be used to improve the matching accuracy. Experimental results show that our system can track vehicles in a wide area with satisfactory precision.