一种基于车载信号还原机动车道3D地图的大数据方法

提出一种基于多台行驶中汽车产生数据重建机动车道3D地图的大数据策略.每台在线汽车上的程序实时上传经过优化的汽车3D坐标信息至服务器.优化方法为使用最小二乘法结合卡尔曼滤波器，利用汽车总线信号实时修正汽车的位置，相比GPS信号，经纬度和高度误差均降低50%以上.此外，还使用遗传算法代替卡尔曼滤波器，进一步降低卡尔曼滤波器的经纬度误差达16%.其次，服务器根据来自多台在线汽车上传的数据建立道路表面的3D点云数据库，并使用K-聚类算法进行数据挖掘，可推算出具有多条行车线道路的每条行车线的中心轨迹，以此建立机动车道3D地图.所建立的地图可为汽车能耗优化策略提供数据支持，降低行驶能耗.

A Big Data Method to Rebuild 3D Road Map Based on Vehicle Data

A big data based 3D map reconstruction method was presented based on multiple vehicles data. On individual vehicle site， in the proposed algorithm， the least square method and Kalman filter were used， and the refined vehicle instant 3D position was uploaded to the server. The original GPS signal error was reduced by over 50%. Next， genetic algorithm was used instead of Kalman filter， resulting in the error being further reduced by over 16%. On the server site， a 3D road surface point cloud database was generated based on the data from multiple vehicles. K-means method was used as the data mining strategy to search for the lane centers from roads with multiple lanes. The reconstructed map can be used by every online vehicles that support the relevant researches for the optimum driving strategy.