基于视频图像与直接线性变换理论的车辆运动信息重构

Vehicle Movement Information Reconstruction Based on Video Images and DLT Theory

运用摄影测量中的直接线性变换理论,并根据事故现场的标记点信息,解得正投影变换的二维直接线性变换系数。通过对事故监控视频进行解帧得到相片组,在解帧相片组上标记车轮与道路的接触点,并对车轮标记点进行正投影。由标记点拟合出轨迹曲线,再通过积分求得轨迹曲线的长度,结合视频的帧频信息,进而求得肇事车辆的车速等。以一起真实交通事故为例,运用该方法,求得其运行轨迹长度和车速等数据。

By applying direct linear transformation theory in photogrammetry and based on the marked points information in accident site, the 2D direct linear transformation coefficients of orthographic projection transformation are solved out. By frame decoding of accident surveillance video, a group of pictures are obtained, and then a series of contact points between wheel and road are marked on decoded picture group and an orthographic projection is conducted on the marked points on wheel. Based on these marked points the trajectory curve is fitted, and the length of trajectory is calculated by integration. Combined with frame rate information of video the speed of accident vehicle is finally obtained. The method is applied to a real traffic accident case with the data of driving trajectory length and vehicle speed calculated.