| Abnormal trajectory detection algorithms often ignore the quality of trajectory data,and frequently rely on popular routes or determined reference routes during detection,which will spend a lot of time on calculating a certain distance between trajectories and also reduces the accuracy of detection results;With the continuous increase of trajectory data,offline abnormal trajectory detection can screen a large number of trajectory sets,detect abnormal trajectories,and obtain a more complete trajectory set for relevant applications,but it cannot detect newly generated anomalies,which has great limitations in daily applications.To solve the above problems,this paper proposes an anomaly trajectory detection method based on grouping map matching,which improves the accuracy of anomaly trajectory detection.On this basis,an incremental anomaly trajectory detection algorithm based on trajectory extension is proposed,which breaks through offline detection and achieves incremental trajectory anomaly detection.The main research contents are as follows:(1)An anomaly trajectory detection algorithm SDATD based on grouping mapping is proposed.Firstly,divide the trajectories adjacent to the starting points and ending points into the same group,and all trajectories will be calculated in groups for subsequent calculations;Secondly,a multi-scale comprehensive mapping method of trajectory points is proposed.which adopts the idea of grouping matching to carry out comprehensive mapping of trajectory points,prunes invalid trajectory points according to the characteristics of grouping,speeds up the connection speed of mapped trajectory points,preserves the characteristics of each group of trajectories,and improves the quality of data and groups.Then,a similar trajectory sequence is constructed to find representative trajectories through the directional cost sub trajectory similarity search;Finally,the trajectory points are reduced based on the driving conditions at the intersection,and the abnormal threshold value is calculated using the reduced trajectory.The abnormal trajectory is determined by comparing the threshold value and the distance between the trajectory and the representative trajectory.(2)Based on the above research,an incremental anomaly trajectory detection algorithm SDOATD based on trajectory extension is proposed.Firstly,map mapping the new trajectory and find its corresponding trajectory group;Secondly,a fixed s-d trajectory extension method is proposed to calculate the trajectory consumption and extend the newly added trajectory to the corresponding destination efficiently through this algorithm,so as to obtain the complete trajectory and effectively improve the quality of the trajectory data;Then calculate the similarity and deviation between the new trajectory and the reference trajectory through the similarity measurement of the trajectory,and then extend the new trajectory to the reference trajectory;Finally,calculate the deviation between the reference route and the optimal extended route to obtain the anomaly threshold,and detect the deviation between the current trajectory and the extended trajectory at each time stamp,and judge the anomaly of the trajectory by comparing with the threshold.Experiments were conducted on real taxi data sets.The efficiency of SDATD is verified by comparing SDATD with three traditional anomaly trajectory detection algorithms;By comparing SDOATD with two incremental anomaly detection algorithms and comparing the results under different incremental trajectories,the efficiency and accuracy of the SDOATD method are verified. |
