基于轨迹数据的道路交叉口及其结构提取方法

Road intersections and structure extraction method based on trajectory data

针对轨迹大数据自动提取城市道路,特别是交叉口问题,该文提出一种新的基于出租车轨迹数据的道路交叉口识别及其结构提取方法:识别轨迹中的转弯片段及其中心点;利用中心点周围轨迹的面积自适应确定聚类半径,并借助自适应聚类半径的DBSCAN方法识别可能的交叉口;对交叉口的轨迹数据进行清洗,用公共子轨迹相似性和角度相似性对转弯片段聚类;利用聚类簇删除非交叉口并提取道路交叉口中心线,获得详细的道路交叉口结构。以北京市出租车轨迹数据为实验数据,不仅识别出了相距较近的道路交叉口,也识别出了复杂的道路交叉口,并且得到了道路交叉口的详细结构。实验结果表明本文所提方法是可行的。

Automatic extraction of urban roads,especially intersections,based on big trajectory data has become a research hotspot.In this paper,a new method of road intersections detection and structure extraction based on taxi trajectory data is proposed.Firstly,the turning segments and their center points are identified.Secondly,the area of the trajectory around the center point is used to determine the cluster radius adaptively,and the DBSCAN method of the adaptive clustering radius is used to recognize possible intersections.Then,the trajectory data cleaning of the intersections is done,the turning segments are clustered by common sub-trajectory similarity and angle similarity.Finally,the clustering clusters are used to delete the non-intersection and extract the central lines of the road intersection to obtain detailed structures of road intersections.Taking Beijing taxi trajectory data as experimental data,not only the near road intersections,but also the complex road intersections were identified,and the detailed structure of the road intersections was obtained.These experimental results show that the proposed method is feasible.