基于浮动车轨迹数据的路口车道结构求解  

Study of urban intersection lane structure based on floating car tracking data

作　　者：杨林[1,2] 左泽均[1] 叶亚琴[1] 周顺平[1] 安晓亚[2,3] 

机构地区：[1]中国地质大学(武汉)信息工程学院,武汉430074 [2]地理信息工程国家重点实验室,西安710054 [3]西安测绘研究所,西安710054

出　　处：《测绘科学》2014年第12期76-80,共5页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41201385;41201469;41301426;41301427;41371422);地理信息工程国家重点实验室开放基金(SKLGIE2013-M-4-5);武汉市科技计划项目(2013010501010123);中央高校基本科研业务费专项资金项目(CUGL120274)

摘　　要：文章针对道路交叉口处车道转向信息变化快、采集困难的问题,提出了利用浮动车轨迹数据自动提取道路交叉口处车道结构及转向的方法。首先根据选取规则对研究区域轨迹点进行约束选取,并依据预处理策略对研究区轨迹点序列进行去噪与优化;其次运用混合高斯密度模型对交叉口处道路的细部车道结构进行建模,利用贪心EM算法解算高斯混合密度模型参数,求得车道数目和车道结构特征;然后给出通过轨迹相关性计算车道转向的计算思路;最后选取一组真实路口数据进行实验,结果验证了该方法的有效性:该方法为车道信息更新提供了参考,亦可应用于精细化导航以提高出行者的服务体验。For road intersection, the turning information of lanes changes quickly and it is difficult to be gathered, therefore, the method of extracting lane structure and turning information using the floating car data was proposed in the paper. Firstly, the trajectory points set were selected by the selection rule and processed by optimized rules; secondly, Gaussian mixture density model was used for lane structure mod- eling of intersection and greedy EM algorithm was used for calculating the parameters to get the lane num- ber and the lane structure; thirdly, the strategy of calculating the lane turning was given by the relation between the trajectory and the lanes. Finally, a group of real intersection data were selected for experiment, and the results demonstrated the effectiveness of this method, which could not only serve the lane information update, but also be applied in finer navigation system for the improvement of the navigation experience,

关 键 词：浮动车轨迹 车道行驶方向 高斯混合模型 

分 类 号：P208[天文地球—地图制图学与地理信息工程]