基于视频车流轨迹的虚拟车道划分方法  

Virtual lane division method based on video traffic trajectory

作　　者：刘继聪 陈庆奎[1] LIU Ji-cong;CHEN Qing-kui(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机工程与设计》2020年第4期1174-1180,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61572325、60970012);高等学校博士学科点专项科研博导基金项目(20113120110008);上海重点科技攻关基金项目(14511107902、16DZ1203603);上海市工程中心建设基金项目(GCZX14014);上海市一流学科建设基金项目(XTKX2012);沪江基金研究基地专项基金项目(C14001)。

摘　　要：针对车道线磨损、临时改道以及非结构化道路等情况下的车道划分问题,在利用YOLOv3得到车辆检测模型前提下,提出基于视频车流轨迹的虚拟车道划分方法。密度矩阵统计时间t内由车辆检测模型得到车流量密度分布,运用三维坐标系对其进行分析;使用EM算法对一元混合高斯模型求解;建立虚拟车道宽度数学模型,运用3σ准则得到车道边界点集合,利用最小二乘法对边界点进行曲线拟合,完成虚拟车道线划分。该方法可以有效避免环境和天气因素对车道线检测的影响,具有一定的鲁棒性和灵活性。实验结果表明,该方法在不同道路中能够取得88.7%的准确率。Aiming at the lane division problem under the condition of lane line wear,temporary diversion and unstructured roads,the virtual lane division method based on video traffic trajectory was proposed on the basis of using YOLOv3 to obtain the vehicle detection model.The vehicle density distribution was obtained from the vehicle detection model in the statistical time t of the density matrix,and the three-dimensional coordinate system was used to analyze it.The EM algorithm was used to solve the unary hybrid Gaussian model.The virtual lane width mathematical model was established.The 3σwas used to obtain the set of lane boundary points,the least square method was used to curve the boundary points to complete the virtual lane line division.The method effectively avoids the influence of environmental and weather factors on lane line detection,and it has certain robustness and flexibility.Experimental results show that the method can achieve an accuracy of 88.7%on different roads.

关 键 词：虚拟车道 车辆检测 密度矩阵 混合高斯分布 3σ准则 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]