交叉口机动车运动轨迹特征提取与标定

Extraction and Calibration of Trajectory Characteristics of Vehicles at Intersections

为探讨机动车在交叉口的运行特性,采用复合特征提取算法获取图像上机动车运行的轨迹特征;在多边形线性扫描算法的基础上,考虑摄像机成像畸变的影响,引入中心偏移因子,提出了考虑中心偏移的多区域扫描标定算法,将运行轨迹图像特征转化为真实的运动特征;最后,与多边形线性扫描算法的计算结果及实测数据进行了对比,结果表明:该算法能够有效地提取交叉口机动车的运行轨迹,准确地表征机动车在交叉口的相关运行特性;与实测车速相比,计算得到的机动车速度误差小于4%.

In order to probe into the running characteristics of vehicles at intersections,the composite feature extraction algorithm was used to obtain the trajectory characteristics of vehicles in images.Based on the algorithm of polygon linear scan calibration,considering the influence of the distortion of CCD（charge-coupled device） camera and introducing center deviation factor,an MRSCCO（multi-region scanning calibration considering about center offset） algorithm was proposed to transform the trajectory features of images into real motion features.Finally,data extracted using this algorithm were compared with both results obtained by the algorithm of polygon linear scan calibration and real data.The experimental results show that the proposed algorithm can effectively extract the trajectories of vehicles at intersections and accurately express the relevant characteristics.Compared with the real vehicle speed,the vehicle speed obtained by the MRSCCO algorithm has an error of less than 4%.