基于背景差分法的交通事件智能检测系统

为快速、准确地检测出交通事故的出事地点,设计了基于前后背景差分的智能检测系统。通过研究一段时间内静态位置上的像素变化,并选取合适的阈值衡量变化程度,从而检测是否存在交通事件。在选取阈值方面,结合了运动目标分割中的背景减法和OTSU算法,并根据相邻帧间的相关性对原算法加以改进,获取了自适应效果较好的分割阈值,从而实现事故的智能检测。截取浙江省杭甬高速公路的交通视频图像进行仿真,结果显示,该系统达到了运行数据量少,检测速度快,准确性高的效果。

Intelligent Traffic Event Detection System Based on Background Difference

An intelligent detection system based on the difference between the former and latter background was designed to rapidly and accurately detect the scene of traffic accident.The pixels′ change of static positions in a period of time was investigated while selecting an appropriate threshold to weigh the degree of change;whether the traffic incident took place was then detected.The moving object segmentation in background subtraction was combined with OTSU algorithm when the threshold was determined.The original algorithm was modified with the correlation between two adjacent frames;a better self-adaptive threshold was then obtained,which was beneficial to achieve intelligent incident detection.The video of Hangyong Expressway in Zhejiang Province was used to simulate the traffic event detection.Accurate results were achieved in high test speed with a small quantity of operating data.