摘要
针对目前在遗留物检测方面所存在的检测范围小、误检率过高和无法捕捉到遗留物放置者等问题,设计了一种基于全方位计算机视觉的遗留物检测系统;首先,采用全方位视觉传感器(Omni-Directional Vision Sensors,ODVS)来获得更大范围的全景视频检测区域;其次,利用一种基于两个不同更新率的改进的混合高斯分布模型的建模方法,获得两个背景模型,再通过当前帧与获得的两个背景模型进行差分运算得到当前帧的暂时静止对象;再次,根据时间指标和距离指标判定暂时静止对象是否属于遗留物;最后,将遗留物所处的空间展开成透视图来进行报警;实验结果表明,该系统能有效地检测全景范围内的遗留物,具有较高的检测精度和鲁棒性。
At present, video surveillance used in the field of abandoned objects detection has faced the problems： limit of detection range, high mistake of detection, unable to detect the owner of abandoned objects. Based on this situation, designed an abandoned objects detection system based on computer vision. At first, Omni--Directional Vision Sensors （ODVS） is used to get larger detection area. Secondly, we use an improved mixed Gaussian modeling method based on two different update rates in order to get two background models, then subtract cur- rent frame from the two background models in order to detect the temporarily static objects. Thirdly, judge the temporarily static objects as abandoned objects or not, using the standard of time and distance. At last, alarmed after the abandoned object was detected. Experimental results show that the system is robust enough to detect the abandoned objects effectively.
出处
《计算机测量与控制》
CSCD
北大核心
2010年第3期517-519,523,共4页
Computer Measurement &Control
