To improve the performance of the traditional map matching algorithms in freeway traffic state monitoring systems using the low logging frequency GPS (global positioning system) probe data, a map matching algorithm ...To improve the performance of the traditional map matching algorithms in freeway traffic state monitoring systems using the low logging frequency GPS (global positioning system) probe data, a map matching algorithm based on the Oracle spatial data model is proposed. The algorithm uses the Oracle road network data model to analyze the spatial relationships between massive GPS positioning points and freeway networks, builds an N-shortest path algorithm to find reasonable candidate routes between GPS positioning points efficiently, and uses the fuzzy logic inference system to determine the final matched traveling route. According to the implementation with field data from Los Angeles, the computation speed of the algorithm is about 135 GPS positioning points per second and the accuracy is 98.9%. The results demonstrate the effectiveness and accuracy of the proposed algorithm for mapping massive GPS positioning data onto freeway networks with complex geometric characteristics.展开更多
针对停车场视觉建图的回环检测方法对目标级别的语义信息关注较少、在长时间大范围的建图过程中地图一致性与持久性较差的问题,设计了一种语义回环检测算法。该算法基于语义实例,使用图匹配方法找到回环帧并进行帧间位姿估计,生成回环...针对停车场视觉建图的回环检测方法对目标级别的语义信息关注较少、在长时间大范围的建图过程中地图一致性与持久性较差的问题,设计了一种语义回环检测算法。该算法基于语义实例,使用图匹配方法找到回环帧并进行帧间位姿估计,生成回环约束。在同准确率下,该回环检测算法的召回率均高于基于ORB(oriented fast and rotate brief)描述子和词袋法的回环检测方案。在停车场建图与定位试验中,建图轨迹与轨迹真值的绝对误差均小于1 m,定位误差均小于0.3 m,满足对应的技术要求。试验结果表明,本文提出的语义回环检测算法的回环检测性能优于传统回环检测算法,适用于停车场视觉建图任务。展开更多
文摘To improve the performance of the traditional map matching algorithms in freeway traffic state monitoring systems using the low logging frequency GPS (global positioning system) probe data, a map matching algorithm based on the Oracle spatial data model is proposed. The algorithm uses the Oracle road network data model to analyze the spatial relationships between massive GPS positioning points and freeway networks, builds an N-shortest path algorithm to find reasonable candidate routes between GPS positioning points efficiently, and uses the fuzzy logic inference system to determine the final matched traveling route. According to the implementation with field data from Los Angeles, the computation speed of the algorithm is about 135 GPS positioning points per second and the accuracy is 98.9%. The results demonstrate the effectiveness and accuracy of the proposed algorithm for mapping massive GPS positioning data onto freeway networks with complex geometric characteristics.
文摘针对停车场视觉建图的回环检测方法对目标级别的语义信息关注较少、在长时间大范围的建图过程中地图一致性与持久性较差的问题,设计了一种语义回环检测算法。该算法基于语义实例,使用图匹配方法找到回环帧并进行帧间位姿估计,生成回环约束。在同准确率下,该回环检测算法的召回率均高于基于ORB(oriented fast and rotate brief)描述子和词袋法的回环检测方案。在停车场建图与定位试验中,建图轨迹与轨迹真值的绝对误差均小于1 m,定位误差均小于0.3 m,满足对应的技术要求。试验结果表明,本文提出的语义回环检测算法的回环检测性能优于传统回环检测算法,适用于停车场视觉建图任务。