面向多区域视频监控的运动目标检测系统

摘要

In order to realize multi-regions moving obj ect detection in video surveillance,the shortage of traditional moving obj ect detection algorithm is analyzed.Combining the background subtraction al-gorithm and frame difference algorithm,we propose a method which dynamically refresh the back-ground.This algorithm can adapt with the background automatically,and reduce error detections caused by the change of background.We build a video surveillance system based on a FPGA chip,and implement the algorithm on the chip.The system can detect moving object in a 640 pixel×480 pixel, 30 frames/s video stream.The system also provides a function which detects moving object in inde-pendent regions.The number,size and location of the regions can be set easily by simple keystrokes. The system’s configuration is convenient,and has strong flexibility.It can be utilized in various fields such as intelligent surveillance and production safety.%为了实现视频监控现场多区域运动目标检测,分析了传统运动检测算法的不足,结合帧间差分法和背景差分法,提出背景动态更新的运动检测算法。该算法能自适应背景的变化,减少由背景变化造成的误检测。构建基于 FPGA 的视频监控系统,在 FPGA上用该算法实现了640 pixel×480 pixel,30帧/s 视频信号流的运动目标实时检测。系统提供了分区域运动目标检测的功能。检测区域的大小、位置和个数可通过简单的按键操作进行设定。测试结果表明,系统可以实时地对进入划定区域的运动目标进行检测和闪烁告警,且资源占用较少,适合在小规模的 FPGA上进行实现。