一种非重叠多摄像头的实时监控系统

大范围场景的监控需要使用多个摄像头。论文利用运动目标的颜色信息和路径特征,设计了一种非重叠多摄像头的实时监控系统。系统采用分布式多层次结构,在进行单摄像头层的处理时,根据像素点亮度变化检测和跟踪运动目标,同时获取运动目标的外形信息和路径特征;在进行多摄像头层的处理时,使用估计目标外形变化和建立路径模型方法融合多个摄像头信息,实现目标在非重叠多摄像头的跟踪。该系统不要求校准摄像头,也不要求建立完整的场景模型,即便在有亮度变化的环境中,仍能立即准确跟踪目标。实验证明提出的方法有好的跟踪效果。

A Realtime Surveillance System for Disjoint Multiple Cameras

Multiple cameras are required in the surveillance of wide areas.In this paper,a realtime surveillance system based on disjoint multiple cameras is designed.The system uses distributed architecture with many layers.The system detects and tracks moving objects using intensity cues at the single-camera layer.With color information and path features of moving objects their appearance changes are estimated and path models are established in multiple cameras. At the multi-camera layer,the system uses data fusing algorithm to realize object tracking based on disjoint multiple cameras.Compared with other surveillance system,this system tracks objects accurately and immediately in light change cameras with disjoint views,without requiring explicit calibration.Experiment results validate the proposed approach.