视频图像的车辆速度实时检测算法研究

Vehicle speed detection algorithm research of real-time video image

在监控视频图像中因拍摄角度的问题,导致图像中的车道线并不是竖直的。针对在此图像中选取对应块,实现车辆速度实时检测的问题,分析了车身的特征及图像中车道线的倾斜角度,选择车灯作为对应块,选取车灯存在的候选区域;然后根据车灯的对称性强度筛选车灯带,实现车灯的准确定位,并把车灯移动的距离通过摄像机标定转换到实际坐标中,从而实现车辆速度的实时检测。实验结果证明,该方法运行速度快、定位准确,因此可以广泛地应用于红绿灯路口环境中,实现视频图像的车速实时检测。

The lane line in the video, due to the shooting angle, is not vertical, thus leads to a difficulty to detect the real- time speed of cars passing by taken areas in the image. With regard to this problem, this paper analysed the features of the car and the inclination angle of the lane line in the image, and selected the car headlight as the reference area to define the candi- date area which was existed the car lights. Then, the paper selected car lights zone according to the symmetry of light intensity to realize the pinpointing of the car lights, and with the camera calibration, transformed the move distance of car headlight into world coordinate, leading to achieve real-time speed detection of passing cars. The results show that this is fast method with accurate locating abihty. So this algorithm can be wide[y used in realizing the real-time speed detection in video image of traf- fic light intersection.