Research On Traffic OD Flow Oriented Crowd Movement Pattern Mining And Visualization

With the development of wireless sensor technology,communication technology and positioning technology in recent years,a large amount of traffic OD flows can be obtained.The traffic OD flows contains the commuting rules and behavior patterns of moving residents.Mining and analyzing these traffic OD flows can provide intelligent information services and decision support for urban road planning,public transportation optimization,and urban functional area division.The geospatial interaction of traffic OD flow data has been paid much attention by scholars at home and abroad,and a lot of research results have been obtained.However,there are still some problems.For example,due to the large volume of traffic OD flow data and the relatively complex structure,the existing mining algorithm has high time complexity,low effectiveness and difficulty in completing the pattern mining of real-time traffic OD flow and multi-density traffic OD flow.In addition,existing research has not fundamentally solved the problem of the visualization of movement patterns between zones.This dissertation mainly studies algorithms for mining movement patterns between zones based on traffic OD flows and visually analyzes the results of the pattern mining.The main innovation contributions of the dissertation are as follows.1.In order to solve the problem of high time complexity and low efficiency of existing algorithms for dealing with traffic OD flow with spatio-tempral attributes,algorithm DBMPZ(Density-Based Movement Patterns between Zones)is proposed to mine spatial clusters of traffic OD flows.Algorithm DBMPZ introduces the concept of density,which makes the algorithm more robust to noise.The relationship between the origin zone and destination zone of DBMPZ pattern is close,which can better reflect the interaction between zones.And algorithm STMPZ(Spatiao-Temporal based Movement Patterns between Zones)is proposed to mine spatio-temporal clustes of traffic OD flows.Algorithm STMPZ can quickly search the neighbors of the traffic OD flow by using the spatio-temporal index.In the process of pattern mining,the spatial and temporal similarity between traffic OD flows are considered at the same time,thereby achieving the integration of spatio-temporal characteristics.In addition,in the measurement of spatial similarity,the constraints of the road network are considered,which makes the result more accurate and realistic.Finally,the real-world and synthetic datasets are used to verify algorithm DBMPZ and algorithm STMPZ.The experimental results show that the two algorithms are better than algorithm MZP and algorithm SSTFC in terms of efficiency and effectiveness.2.In order to mining movement patterns of real-time traffic OD flows,algorithm RTMPZ(Real-Time-based Movement Patterns between Zones)is proposed to mine the real-time clusters of traffic OD flows.In the process of mining,the dual time windows are designed to solve the problem that the traffic OD flows cannot have corresponding position information at each timestamp and the time window can not slide.Moreover,algorithm RTMPZ only need the data within the dual time windows to update the existing patterns,which improves the efficiency of the algorithm.Finally,the synthetic dataset and the Shanghai taxi commuting dataset are used to verify the algorithm.The experimental results show that algorithm RTMPZ has the ability to process real-time Traffic OD flow efficiently.3.In order to mining movement patterns by the traffic OD flow with different densities,algorithm MDMPZ(Multi-Density-based Movement Patterns between Zones)is proposed to mine spatial clusters with multi-density by introduced a new similarity measurement between traffic OD flows.The similarity includes two parts: the similarity of the neighbor distribution and the similarity of their shared neighbors.Based on the similarity measurement,algorithm MDMPZ can mine movement patterns between zones with different sizes,different densities,and different shapes.Finally,the synthetic dataset and the Shanghai taxi commuting dataset are used to verify algorithm MDMPZ.Experiment results show that algorithm MDMPZ can mine the movement patterns between zones with different densities quickly and effectively.4.In order to solve the problem that there are many characteristics of movement pattern with spatial characteristics,and different patterns may have spatial overlap,a novel visualization method of movement patterns between zones based on the concept of geographic information maps is proposed.For visualizing a single movement pattern,in order to reflect the contribution of different zones in a pattern,some visual variables such as color and size are used to express the multiple aspects at the same time.For visualizing multiple movement patterns appearing at the same location,in order to solve the problem of overlapping,the multiple movement patterns can be divided into several categories,such as one-to-one,one-to-many,many-to-many,one-way or two-way movement and visualize them respectively.5.The movement pattern with spatio-temporal characteristics includes multiple temporal features,spatial features,and spatio-temporal composite features.In order to be able to present these features comprehensively and quantitatively through visual variables,and to make the visualization results concise and intuitive,a visualization method of movement patterns between zones by introduce the concept of spatiotemporal interactive prism is proposed.This method can effectively extract various features,such as temporal features,spatial features,regional features,and spatiotemporal features of the movement patterns,and visualize them in a geometrical way.Comprehensive expression of the movement patterns features can help to understand the relationship between the various features and the rules implied in the movement patterns intuitively and deeply.Based on the theory of spatial interaction in geographic information science,this dissertation proposes a movement pattern mining algorithm framework for traffic OD flows with different characteristics,and puts forward the corresponding visualization methods.This study is not only a supplement to the spatial interaction analysis model,but also an extension of spatial interaction visualization method,which provides a method support for the analysis of regional correlation under the influence of traffic OD flow.