A real-time trajectory co-movement motion mode detection method based on flink

Abstract

The invention discloses a Flink-based real-time trajectory co-movement motion mode detecting method, and the steps of the method are as follows: (1) collecting real-time trajectory data of a certain period of time in the application to obtain sample data; (2) discretizing sample data according to the timestamp to obtain multiple snapshots; (3) using each snapshot obtained in step (2), constructinga GR-index for the data, and obtaining a corresponding spatial division; (4) in the clustering stage according to the space division obtained in the step (3), querying and clustering the range of thedata by using the DBSCAN method; (5) in the enumeration phase, each snapshot after the cluster obtained in the step (4) is used. (5) In the enumeration phase, using the id-based partitioning technique for each snapshot after the clustering obtained in step (4), real-time enumeration, and the co-movement motion mode in which each time slice meets the constraint condition is output. The present invention greatly improves the efficiency of motion mode detection and provides optimal performance.

Description

technical field [0001] The invention relates to trajectory data mining technology in the field of computer big data, in particular to a co-movement motion pattern detection method based on Flink and real-time trajectory data. Background technique [0002] With the popularity of positioning devices, a large number of trajectories are continuously generated from various devices in the form of time-space sequence records. Different from static data, real-time trajectory data is a data sequence continuously generated by multiple moving objects. At present, the analysis technology for static trajectory data has become mature, and the analysis technology for real-time trajectory data has become a research hotspot due to its complexity and importance. [0003] Apache Flink is an open source system for processing stream data and batch data. The initiators of Apache Flink believe that regardless of batch data or stream data, many practical application scenarios about them can be exp...

AI Technical Summary

Problems solved by technology

Existing research on trajectory motion pattern detection only focuses on historical data. However, these offline algorithms can no longer meet the needs of real-time analysis of streaming data, because the problem definition for historical static data cannot be directly transferred to the real-time environment.

In addition, for real-time trajectory data, the research problems are mainly simple range query and nearest neighbor solution, and there are many gaps in the work of analyzing motion patterns of real-time trajectory data in a distributed framework

Images