A trajectory anomaly detection method based on dual perspectives of time and space

Abstract

The invention discloses a track anomaly detection method under dual perspectives of time and space, which preserves the original local features of the track by cutting the track into multiple track segments; comprehensively considers the abnormality of the track from the two perspectives of time and space , in order to improve the operating efficiency of the method: introduce the concepts of time period and space grid, and obtain the trajectory segments that are similar to the time period and space grid; the present invention uses the number of abnormal trajectory segments in the trajectory to account for the total trajectory segments In order to reduce the probability of misjudgment of the test trajectory, the test trajectory is added to the original trajectory movement pattern, and the trajectory data is retrained to obtain a new trajectory movement pattern and judge the existing trajectory again. Whether the trajectory determined to be abnormal is still an abnormal trajectory relative to the new trajectory movement mode.

Description

technical field [0001] The invention belongs to the technical field of track data detection, and in particular relates to a track abnormality detection method under dual perspectives of time and space. Background technique [0002] With the rapid development of communication technology, sensor network technology and positioning technology, as well as the popularization and application of GPS positioning equipment, mobile smart terminals and other equipment, a large amount of individual trajectory information is collected and stored in the database. Trajectory information often hides the law of movement of individuals. If a certain law of movement can be excavated from known trajectory information and used to detect abnormal trajectories, it will be of great significance to the fields of individual abnormal behavior detection, urban planning, and natural disaster prediction. . [0003] At present, there are two main categories of trajectory anomaly detection methods. One is...

AI Technical Summary

Problems solved by technology

The disadvantage of these methods is that the difference in the direction of individual trajectory movement is not considered when dividing trajectory segments.

The disadvantage of this method is that the trajectory is mapped so that the acquired feature points are not the original trajectory points in the trajectory, and the original motion information of the trajectory is ignored.

However, this type of detection method only utilizes the motion information of feature points in the detection process, while the motion information of some trajectory points is ignored, so this type of detection method ignores the local characteristics of the trajectory.

[0005] After analyzing the current abnormal trajectory detection method, it is found that there are three main problems: (1) the method of using the abnormal degree of the trajectory point to judge the overall abnormality of the trajectory has a high time complexity; (2) through The detection method of obtaining feature points in the trajectory to judge the abnormality of the trajectory, although the time complexity of the method is reduced, but it often obtains the feature points according to the difference in position and speed, and does not consider the difference in the direction of motion of the trajectory points and ignores the The local features of the trajectory, such as figure 1 As shown, A is the starting point of a certain trajectory, B is the end point of the trajectory, C and D are two adjacent trajectory points in the trajectory, it can be seen from the figure that the directions of the speeds of C and D are opposite, if Their speeds are equal, and it may not be possible to distinguish the difference between the motions of points C and D without considering the motion direction of the trajectory point; (3) In most methods, the motion pattern of the trajectory will not change once it is established, but the trajectory data Sequential and variable, with movement patterns that may change over time, such as figure 2 As shown, for a trajectory time series data, it may be from figure 2 During the period from trajectory ① to trajectory ⑤, the trajectory has a relatively stable motion pattern, and from a certain moment, that is, trajectory ⑥ in the figure, the motion pattern of the time series trajectory data has changed. The existing anomaly detection method Trajectories ⑥, ⑦, ⑧, and ⑨ will be judged as abnormal trajectories, but in fact, these trajectories may be normal as time goes by

Images