Abstract: To address the issue of inefficiency and inconvenience in camera's manual calibration required by the existing video-based vehicle speed detection, a robust algorithm was proposed to carry out automatic calibration of typically-configured road-monitoring video camera, including calibration of its focal length, pitch angle and distance to the road surface. The relationship between vanishing points formed by two groups of orthogonal lines was used to calibrate the focal length and pitch angle of the camera. Then the statistical width data collected from multiple vehicles detected in video stream was used to extract the distance between the camera and the road surface. The experimental results show that the proposed algorithm has high efficiency, excellent accuracy and high reliability. The proposed algorithm can be deployed as an effective automatic camera calibration way in helping the existing traffic monitoring video devices to carry out automatic collection, analysis and monitoring of vehicle speed, vehicle type, traffic flow data and enforcing traffic speed limit law.

Key words: real-time vehicle speed detection, automatic vanishing point detection, automatic camera calibration, Intelligent Transportation System (ITS), video processing

摘要： 针对目前基于视频的车辆测速方法均需通过手工标定而造成的低效和可操作性差的问题，提出了一种对典型配置的道路监控摄像机的焦距、俯仰角、离地距离等重要参数进行自动标定的方法。首选利用自然场景中两组正交平行线在视频图像中形成的消失点之间的内在关系对摄像机的焦距和俯仰角实施精确标定；在此基础上利用视频中目标车辆群体的平均宽度对摄像机与地面之间距离进行自动标定。实验表明，该算法具有参数测量精度高和可靠性好等优点，可作为现有道路视频监控设备实施车辆速度、类别、流量等数据的自动采集、分析和监控，以及电子违章抓拍设备的有效自动标定手段。

关键词: 实时车辆速度检测, 消失点自动检测, 摄像机自动标定, 智能交通系统, 视频处理