集成时空邻近与卷积网络车道级高精度定位算法  

Lane-level high-precision positioning algorithm based on integrated spatio-temporal proximity and CNN

作　　者：滕文鑫 张建辰 邵杰 TENG Wenxin;ZHANG Jianchen;SHAO Jie(College of Geography and Environment,Henan University,Kaifeng 475004,China;State Key Laboratory of Mapping and Remote Sensing Information Engineering,Wuhan University,Wuhan 430070,China;Key Laboratory of Urban Land Resources Monitoring and Simulation,Ministry of Natural Resources,Shenzhen 5180384,China;Henan Industrial Technology Academy of Spatio-Temporal Big Data,Henan University,Zhengzhou 450000,China;Hongluosi Middle School,Capital Normal University,Beijing 101400,China;Henan University key Laboratory of Geospatial Technology for the Middle and Lower Yellow River Regins,Ministry of Education,Kaifeng 475004,China)

机构地区：[1]河南大学地理与环境学院,河南开封475004 [2]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430070 [3]自然资源部城市国土资源监测与仿真重点实验室,广东深圳5180384 [4]河南大学河南省时空大数据产业技术研究院,河南郑州450000 [5]首都师范大学附属红螺寺中学,北京101400 [6]河南大学黄河中下游数字地理技术教育部重点实验室,河南开封475004

出　　处：《测绘通报》2022年第6期1-5,61,共6页Bulletin of Surveying and Mapping

基　　金：自然资源部城市国土资源监测与仿真重点实验室开放基金(KF-2020-05-037)。

摘　　要：提高车道水平定位精度是智能交通系统发展的重要技术之一。本文提出了一种新的车道级定位方法——时空邻近卷积神经网络(STP-CNN),利用时空附近(STP)动态细化候选匹配道路,再进一步采用个性化卷积神经网络(CNN)自适应识别最优匹配车道。该方法通过优化集成GPS、车速和惯性测量单元等参数,实现了厘米级和车道级车辆位置的平滑估计。试验结果验证了该方法的可行性和有效性。Improving the accuracy of lane-level positioning is one of the most important technologies for the development of intelligent transportation system(ITS).We propose a new lane-level positioning method called spatio-temporal proximity and convolution neural network(STP-CNN),in which STP is designed to dynamically refine candidate matching roads,and personalize CNN is further adopted to adaptively identify the optimal matching lane.By optimally integrating parameters such as GPS receiver,driving speed and inertial measurement unit for the new method,centimeter-level and lane-level vehicle position is estimated smoothly.The test results justify the feasibility and efficiency of the proposed method.

关 键 词：卷积神经网络 深度学习 车道级定位 智能交通系统 时空邻近 地图匹配 

分 类 号：U495[交通运输工程—交通运输规划与管理]