基于空间坐标系旋转的高效轨迹匹配算法

Efficient Trajectory Matching Algorithm Based on Spatial Coordinate Rotation

轨迹相似度估计是发现车辆运动特征和轨迹分类的关键,但计算轨迹相似度缓慢,提高轨迹匹配速度可以帮助展开轨迹特征的快速挖掘,因此提出一种基于空间坐标系旋转的高效轨迹匹配算法.首先利用空间坐标系的多次旋转,将轨迹曲线转换成点数等于旋转次数的平均数和方差曲线;然后使用Fréchet距离和皮尔森相关系数衡量平均数曲线间相关性和方差曲线间相关性;最后根据Fréchet平均数、Fréchet方差、皮尔森平均数和皮尔森方差4个参数的大小间接确定原始轨迹间的相似性.在不同的轨迹采样点数目和空间坐标系旋转次数下,基于杭州市出租车轨迹数据,与传统Hausdorff轨迹匹配算法比较轨迹匹配的准确度和速度.基于多次实验的结果表明,该算法保证轨迹匹配准确度的同时,平均可以提高85%的轨迹匹配速度.通过构建包含展示轨迹匹配结果的地图概览、探索轨迹匹配结果差异性的可视化交互组件和选择轨迹匹配参数3大模块的可视化分析系统,探索4种轨迹匹配方法结果的差异性,帮助道路网研究人员更快地匹配真实行车轨迹和寻找相似轨迹群.

Trajectory similarity estimation is the key to discover vehicle motion characteristics and trajectory classification,but the calculation of trajectory similarity is slow,and improving the speed of trajectory matching can help to carry out trajectory characteristics mining fast.Therefore,an efficient trajectory matching algorithm based on the rotation of spatial coordinate is proposed.Firstly,the trajectory curve is converted into the mean and variance curve which the number of points is equal to the number of rotations by multiple rotation of the spatial coordinate system.Then,Pearson correlation coefficient and Fréchet distance are used to measure the correlation of the mean and variance curves.Finally,the similarity between the original trajectories is determined indirectly according to the values of four parameters:Frechet mean,Frechet var,Pearson mean and Pearson var.Based on the taxi trajectory data of Hangzhou,under different trajectory sampling points and spatial coordinate rotation times,compared trajectory matching accuracy and speed with the traditional Hausdorff trajectory matching algorithm.The several experiments show that the algorithm can improve the trajectory matching speed by 85% on average while ensuring the accuracy of trajectory matching.By constructing a visual analysis system that includes three major modules:a map overview showing the results of trajectory matching,several visual interactive components that explore the differences in trajectory matching results,and a module for selecting trajectory matching parameters,explore the differences in the results of the four trajectory matching methods to help road network researchers to match real driving trajectories and find similar trajectory groups faster.