一种利用导航轨迹数据的车道信息提取方法

A Method for Lane Information Extraction Using Navigation Trajectory Data

基于轨迹数据提取车道级道路信息的关键在于道路中心线,然而目前已有研究大都基于先验地图或者粗略估算来获取道路中心线,这严重降低了后续车道信息提取与更新的效率和精度。为此,文章提出了一种利用导航GNSS轨迹数据自动提取车道信息的方法。首先,利用自适应K-means聚类方法对导航轨迹进行方向划分并基于轨迹的密度分布实现轨迹数据清洗;其次,根据机动车轨迹在道路上的位置分布进行拟合,实现车行方向道路中心线的提取;最后,基于约束高斯混合模型实现各个路段的车道数量和车道中心线的提取。实验结果表明,该方法可以准确地提取车道信息,其中车道数量精度为79.8%,车道中心线的位置精度接近1 m,车道宽度精度大都优于0.5 m。

The key of lane information extraction based on trajectory data is the road center line.However,the existing research is based on a prior map or a rough estimate to obtain the center line of the road,so it seriously reduces the efficiency and accuracy of the lane information extraction and update.In this paper,the method of automatic extraction of lane information is proposed by using navigation GNSS trajectory data.Firstly,the method of adaptive K-means clustering is used to divide the direction of navigation trajectories and the trajectory data cleansing is realized based on the density distribution of the trajectory.Secondly,according to the location distribution of the trajectories on the road,the center line of roadway is extracted.Finally,the number of lanes and lane center line are extracted based on the constrained Gaussian mixture model.The experimental results show that the proposed method can extract lane information accurately.The accuracy of lane number extraction is 79.8%,the location accuracy of lane center line is close to 1 m,and the accuracy of lane width is almost better than 0.5 m.