基于约束高斯混合模型的车道信息获取

发布时间：2018-04-13 00:19

本文选题：浮动车数据 + 基本分类器　；参考：《武汉大学学报(信息科学版)》2017年03期

【摘要】：针对现有车道级道路信息获取方法大多存在数据采集成本高、更新周期长、数据处理难度大等缺点,提出了一种基于浮动车数据(floating car data,FCD)的城市车道数量信息快速获取方法。首先根据浮动车数据的空间分布特征,利用Delaunay三角网方法对数据进行优选,通过探测优选后浮动车数据覆盖的宽度间接得到道路宽度;然后将一部分已知车道数量及浮动车数据覆盖宽度的路段作为训练样本,分析其车道数量和浮动车数据覆盖宽度之间的关系构建基本分类器;最后按照待测路段的浮动车数据分布宽度查找基本分类器,获取待测路段可能存在的若干个车道数量类型候选值,并利用约束高斯混合模型对最终车道数量类型进行确认。实验结果表明,该方法实现了从低精度浮动车数据中快速获取车道数量信息,提取精度达到了82.3%。
[Abstract]:In view of the shortcomings of most of the existing methods for obtaining lane level road information, such as high cost of data acquisition, long updating period and great difficulty in data processing, a fast method for obtaining urban lane quantity information based on floating vehicle data floating car dataFCDs is proposed.Firstly, according to the spatial distribution characteristics of floating vehicle data, the Delaunay triangle network method is used to optimize the data, and the road width is indirectly obtained by detecting the coverage width of floating vehicle data.Then, a part of the section with known lane number and floating vehicle data coverage width is used as the training sample to analyze the relationship between the lane number and the floating vehicle data coverage width to construct the basic classifier.Finally, the basic classifier is searched according to the distribution width of floating vehicle data of the road to be tested, and the candidate values of several lane number types are obtained, and the final lane number type is confirmed by using the constrained Gao Si mixed model.The experimental results show that the method can quickly obtain the lane number information from the low-precision floating vehicle data, and the extraction accuracy is up to 82.3%.
【作者单位】：武汉大学测绘遥感信息工程国家重点实验室;深圳大学土木工程学院空间信息智能感知与服务深圳市重点实验室;
【基金】：国家自然科学基金(41271442,40801155,41571430,41671442) 中国航天科技集团公司卫星应用研究院创新基金(2014_CXJJDSJ_02) 深圳市北斗卫星应用工程技术研究中心基金~~
【分类号】：P228.4;U495

【相似文献】