采用轨迹压缩和路网划分的车辆异常轨迹检测

Vehicle Anomalous Detection Using Ttrajectory Compression and Road Network Partition

随着各种网约车平台的蓬勃兴起,网约车犯罪率显著增加,而其行车轨迹往往表现出异常现象.为有效检测存在异常行为的轨迹,提出一种面向道路消耗的车辆异常轨迹检测算法.首先,将建模重点由轨迹数据转移到道路本身,对道路消耗进行建模,同时兼顾时间和距离的影响,有效提高了检测结果的准确性;其次,通过地图匹配概率将轨迹映射到路网空间,有效提高了参与检测的数据质量;然后,依据道路节点和车辆行驶方向是否改变对轨迹进行压缩,减少了内存消耗并提高了算法的效率;第四,提出并定义了消耗阈值矩阵的概念,扩大了算法检测的数据范围;最后,采用真实数据集验证了算法的有效性,并与iBOAT、TRAOD、TADSS和TPRO算法进行对比,验证了本算法具有更高的效率和准确性.

With the burgeoning of car-hailing platforms,the crime rate of online car Hailing increases significantly,where vehicle trajectory often shows abnormal phenomenon.For effectively detecting abnormal trajectories,a vehicle anomalous trajectory detection algorithm for road consumptions was proposed.Firstly,we shift in focus of modeling from trajectories to roads,juggle the influence of time and distance,which effectively improves the accuracy of results.Secondly,trajectories are mapped on road network by map matching probability to improve the data quality.Then,according to road nodes and driving direction,trajectories are compressed to reduce the memory consumption and improve the efficiency of the algorithm.Fourth,we propose and define the concept of consumption threshold matrix,in which data′s range detected is expanded.Finally,the effectiveness of the algorithm is verified by real dataset,and compared with iBOAT,TRAOD,TADSS and TPRO,the algorithm has higher efficiency and accuracy.