

摘 要:停车场出入口是基础地图数据,采用传统人工采集方式时效性差。根据车辆进出地下停车场时GPS信号状态改变的特性,融合DR轨迹、车辆状态、陀螺仪倾角等多传感器数据,提出一种众包地图构建方法,使用DBSCAN聚类及多边形生成算法,挖掘出地下停车场区域范围、出入口及其通行方向。结果表明,该方法提取的出入口数据准确度高,应用于地图数据生产中可实现数据动态更新。导航时将起终点绑定到出入口道路上,可提供最优路线,指引用户进出停车场,改善用户体验。
关键词:车辆轨迹;密度聚类;停车场出入口;多传感器融合;末端导航
中图分类号:U491.7+1;TP311.52 文献标识码:A 文章编号:2096-4706(2020)04-0054-04
Abstract:The gateway to the underground parking lot is basic map data. Its difficult to guarantee data freshness under manual data collection conditions. According to the characteristics of GPS signal state change when vehicles enter and leave the underground parking lot,and the multi-sensor data such as DR track,vehicle state and gyroscope angle are integrated,a method of building crowdsourcing map is proposed. DBSCAN clustering and polygon generation algorithm are used to excavate the area range,entrance and exit of the underground parking lot and its traffic direction. The results show that the accuracy of gateway extracted by using this method is high. Application in map data production can implement dynamic data update. During navigation,the start and end points are bound to the access road,which can provide the best route,guide users to enter and exit the parking lot,and improve the user experience.
Keywords:vehicle trajectory;density clustering;gateway of parking lot;multi-sensor fusion;terminal navigation
0 引 言
有过开车经验的人相信都遇到过找不到停车场入口的尴尬。地图导航软件如何解决末端导航中的用户痛点是笔者工作中的研究课题。究其原因有二:一是导航算路时没有将规划路线的终点绑定到停车场入口道路;二是地图数据中停车场出入口数据覆盖率及鲜度都不足。以上海19Q1数据为例,现有图商数据中,关联了进出口道路的停车场只占总数的28%,且出入口的通行方向经常变化,导致数据准确性不高。传统人工采集停车场出入口数据时效性差,无法反映出入口的后期变化[1]。近年来互联网汽车的兴起讓车辆位置可以发送到云端,通过车辆轨迹大数据挖掘的手段,获取停车场对应的出入口,用全新的方式来补充现有停车场数据的不足。基于停车场出入口数据,改进导航软件中的路径规划算法,实现用户进出停车场时的精确引导。……