基于计算机视觉的汽车流量检测统计

Statistics of vehicle flows based on computer vision

利用计算机视觉和图像处理的方法对交通路口通行车辆进行流量统计.首先,由摄像机视场的标定原理和标定公式,得到了摄像机成像平面内的像素距离与实际三维空间距离的对应关系;然后,采用基于光流场的车辆检测方法,对所有经过设置的虚拟线圈的车辆进行实时检测和信息提取;最后,对提取的车辆信息进行统计和计算,得到实时路况信息.实验结果表明,该算法具有较高的检测概率,能够满足系统的精度和实时处理要求.

On the basis of computer vision and image processing, a method was proposed to improve the statistical results of vehicle flow detection at traffic crossing. According to camera scene calibration theory and expressions, the corresponding relation between the pixels distance on the camera imaging plane and the factual distance of 3D space was obtained. The vehicle detection method based on optical flow was utilized to extract the all vehicles passing the set dummy loop. The real-time route information would be obtained with computing the extracted vehicles. The experimental results showed that this method has higher detection probability and satisfied with system accuracy and robustness require.