摘要
从众源轨迹数据中提取道路几何数据相对于传统的道路数据获取方法具有低成本、高现势性的优点。然而,由于轨迹数据采样稀疏、数据量大、高噪音等特征使得道路中心线提取仍显困难。针对该问题,提出一种基于约束Delaunay三角网的道路中心线提取方法。首先对预处理后的车辆轨迹线构建约束Delaunay三角网,根据整体长边约束准则删除长边以提取道路面域多边形;然后对道路面多边形二次构建Delaunay三角网,提取道路中心线。利用北京市一天时间的出租车轨迹数据进行算法实验,将实验结果与栅格化方法结果进行定性定量地评价分析。结果表明该方法提取的道路中心线数据在几何、拓扑精度方面比栅格化方法提高约10%以上。另外,以复杂环形道路为例,证明了该方法比栅格化方法更适合于复杂道路结构、较大密度差异的轨迹数据。因此,该方法不仅适合大数据处理、结果精度高,且算法成熟、易于实现。
Compared with traditional method,road centerline extraction from crowdsourcing trajectory data offers numerous advantages with respect to labor cost,real-time and data completeness.However,it is difficult to construct road network using big crowdsourcing trajectory data due to the trajectory data with sampling sparse and large data volume and high noise.For this issue,this study tries to explore the question of road centerline extraction by large volume of taxi GPS trajectory data,presenting a new method based on Delaunay triangulation model.The whole method includes three steps.The first one is to pre-process the vehicle trajectory data including the point anomaly removing and the conversion of trajectory point to track line.Secondly,construct Delaunay triangulation within the vehicle trajectory line to detect neighborhood relation.And then,the road coarse polygon is extracted by cutting long triangle edge and organizing the polygon topology.Considering the case that some of the trajectory segments are too long,a interpolation measure is used to add more points for the improved triangulation.Thirdly,construct Delaunay triangulation within the road polygon to extract the road centerline.The centerline is extracted by distinguishing three kinds of triangles and processing the road junction.The experiment is conducted using one day of taxi track in Beijing City.Compared with conventional methods（raster）,experimental results demonstrate that the accuracy of road geometry and topology is improved above 10 percent through the use of the method in this paper.Moreover,the complex ring road is used as a case study to test the proposed method.Experimental results prove that the proposed method is more suitable for complex road structure and trajectory data with different density.As a result,the results achieved with the proposed method show that road centerline can be generated with low cost,high efficiency,good maneuverability,based on crowdsourcing trajectory data,and be very useful for mapping application.
出处
《地理与地理信息科学》
CSCD
北大核心
2016年第3期1-7,共7页
Geography and Geo-Information Science
基金
国家自然科学基金资助项目(41531180)
国家高技术研究发展计划(863计划)资助项目(2015AA1239012)
