论文部分内容阅读
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%。
In order to overcome the shortcomings of high cost of data acquisition, long update cycle and difficulty of data processing, most of the existing methods for acquiring lane-level road information are based on floating car data (FCD) . Firstly, according to the spatial distribution characteristics of the floating car data, the data is optimized by using the Delaunay triangulation method, and the width of the floating car data is indirectly obtained by detecting the width of the floating car data coverage. Then, a part of the road sections covering the known lane numbers and floating car data coverage widths As a training sample, analyzing the relationship between the number of lanes and the coverage width of the floating car to construct a basic classifier; finally, searching the basic classifier according to the distribution width of the floating car data of the section to be measured to obtain the number of lanes that may exist in the section to be measured Type candidate value, and confirms the final lane number type by using a constrained Gaussian mixture model. The experimental results show that this method can quickly obtain the number of lanes from the low-precision floating car data, and the accuracy is up to 82.3%.