基于C-V2X网联车的路侧摄像机标定方法  

Roadside Camera Calibration Method Based on C-V2X Connected Vehicle

作　　者：张长隆 周炜[2] 杜仙童 谢鹏程 崔海涛[2,3] ZHANG Chang-long;ZHOU Wei;DU Xian-tong;XIE Peng-cheng;CUI Hai-tao(Changsha Intelligent Driving Institute Co.,Ltd.,Changsha,Hunan 410006,China;Research Institute of Highway,Ministry of Transport,Beijing 100088,China;Zhonggong Gaoyuan(Beijing)Automobile Testing Technology Co.,Ltd.,Beijing 101103,China)

机构地区：[1]长沙智能驾驶研究院有限公司,湖南长沙410006 [2]交通运输部公路科学研究院,北京100088 [3]中公高远(北京)汽车检测技术有限公司,北京101103

出　　处：《公路交通科技》2024年第9期18-27,共10页Journal of Highway and Transportation Research and Development

摘　　要：为了安全高效地标定公开道路上的路侧摄像机,提出了一种新的标定方法。该方法不需要任何先验知识和封闭道路来阻断交通,克服了现有方法需要人工测量的局限性。它将网联车当作标志物,利用蜂窝车联网(C-V2X)通信技术,通过网联车提供路侧摄像机视野中各位置的WGS-84坐标并采用深度学习算法提取用于标定的网联车图像坐标。对网联车的图像坐标及其WGS-84坐标进行时间戳同步后,基于直接线性变换获取网联车图像坐标与其WGS-84坐标之间的映射关系,完成路侧摄像机标定。在长沙市真实路口进行选址搭建试验平台,采集了82个验证点对本研究方法进行可行性分析试验。结果表明:本研究方法的标定误差基本保持在1.5 m以内,能够满足协作式智能交通系统(C-ITS)应用的车道级定位要求。同时,将该方法与2种具有代表性的路侧摄像机标定方法进行了对比,结果表明本研究方法具有1.80%的最小均方根误差。此外,还探究了网联车姿态、网联车速度和GNSS定位精度对标定精度的影响,分别设计对比试验加以分析,最后依此总结了实用的指导建议,以便采用本研究方法时更容易地获得较高的标定精度。To safely and efficiently calibrate roadside cameras on the open road,the novel calibration method was proposed.The method did not require any prior knowledge or road closures to block traffic,overcoming the limitations of existing methods that required manual measurements.The method treated connected vehicles as markers.The WGS-84 coordinate of each position in the field of roadside camera’s view was provided by the connected vehicle,and the coordinate of connected vehicle image for calibration was extracted by using the depth learning algorithm via cellular vehicle-to-everything(C-V2X)communication technology.After synchronizing the image coordinates and their WGS-84 coordinates with time stamps,based on the direct linear transformation,the mapping relation between connected vehicle image coordinate and its WGS-84 coordinates was obtained to calibrate the roadside camera.An experimental platform was set up at a real intersection in Changsha.There were 82 verification points collected to conduct feasibility analysis experiments.The result indicates that the errors of all 82 validation points are almost no more than 1.5 m,demonstrating that the proposed method can meet the lane-level positioning requirements for the cooperative intelligent transport system(C-ITS)application.Simultaneously,this method is compared with two representative roadside camera calibration methods,and the result shows that the proposed method has the minimum root mean square error of 1.80%.In addition,the influences of connected vehicle attitude,connected vehicle speed and GNSS positioning accuracy on the calibration accuracy were also discussed.The comparative experiments were designed and analyzed respectively.Finally,some practical suggestions were summarized,so that higher calibration accuracy can be easily obtained by using the proposed method.

关 键 词：智能交通 路侧摄像机标定 蜂窝车联网 网联车 WGS-84坐标 

分 类 号：U495[交通运输工程—交通运输规划与管理]