Track data flow-based abnormal moving object detection method

Abstract

The invention discloses a track data flow-based abnormal moving object detection method. The method comprises the following specific steps of 1, measuring abnormality of track data flow; 2, indexing the track data flow; and 3, detecting an abnormal moving object in the track data flow. A track abnormality measurement method considering a spatial position and time information is adopted; an increment updating capability is realized; and lightweight time-space track data flow abnormality measurement is proposed. By indexing a time-space track coverage region, the track data flow index updating frequency is reduced, so that the problem of frequent updating of a track data flow index is radically solved. A framework of organization and management of track data flow and detection query, and aninitialization and query increment maintenance policy capable of meeting the detection real-time requirements of the abnormal moving object in the track data flow are proposed, so that the maintenancecapability of the detection increment of the abnormal moving object is improved.

Description

technical field [0001] The invention relates to a detection method, in particular to a detection method for abnormal moving objects based on track data flow. Background technique [0002] With the maturity of sensor network, global positioning system GPS, wireless communication and other technologies, mobile smart devices with positioning function are widely used, generating a large number of trajectory data streams of moving objects, such as taxi trajectory data streams, animal migration data flow, data flow of personnel movement in large public places, etc. The trajectory data flow is continuously generated every day. For example, the vehicle trajectory is based on the collection of vehicle GPS, with an average daily data volume of tens of millions to billions, and TB-level data trajectory flow information is generated every day; mobile phone trajectory is based on cellular base station sampling, and the daily average data volume One billion to ten billion levels, the tot...

AI Technical Summary

Problems solved by technology

The static trajectory anomaly measurement is to use the Hausdorf distance function to calculate the anomaly degree based on the sub-trajectory segments. This method only considers the geometric shape of the trajectory and does not consider the time dimension, and is not suitable for the trajectory data flow environment; most of the existing trajectory data indexing technologies are For static trajectory data, the sub-trajectory segment is used as the basic unit for indexing, and the trajectory data stream has the characteristics of frequent updates. The update of the trajectory will inevitably cause the update of a large number of sub-trajectories, so that the index is updated frequently, and there are a large number of sub-trajectories between the trajectory segments. Intersection or overlap, the index sub-trajectory segments will cause a large amount of overlap in the index space; the current research on abnormal object detection based on trajectory data is mostly oriented to static trajectory data, the ability of incremental maintenance is relatively weak, and a small amount of abnormal point detection is aimed at a single trajectory data stream Research is only used to detect abnormal sub-trajectory segments in a single trajectory stream, but there are few studies on abnormal moving object detection applied to trajectory data streams, and existing algorithms have certain limitations and deficiencies