Spatio-temporal Trajectory Mining And Application Research Of Regular Moving Objects

In recent years,the rapid development of mobile communication technology,positioning technology and Internet of Things,which not only brought convenience to our lives,but also produced a lot of spatio-temporal trajectory data.The scale of spatio-temporal trajectory data is expanding rapidly,and is developing towards the direction of diversification and diversity.These spatio-temporal data contain a lot of valuable knowledge.However,the types of these spatio-temporal data are varied and complex.The complexity and diversity of spatio-temporal trajectory data also increases the difficulty of data analysis and mining.At present,the key problem for researchers is how to acquire the valuable knowledge from a large number of complex spatio-temporal data.In this paper,the study will focus on trajectories that have unstable and sparse sampling frequencies,and from which to find useful motion patterns.This paper will introduce the research work from two perspectives: theoretical research and application practice,as can be seen in the follow.1.In the existing spatio-temporal data mining methods,there are few researches focus on trajectory data that have unstable and sparse sampling frequencies.In view of this problem,based on the temporal and spatial characteristics of trajectory data,we deeply studies the temporal and spatial structure of trajectories.As a result,we have designed two methods to deal with the spatio-temporal trajectories that have unstable and sparse sampling frequencies.The first method is used to identify and eliminate duplicate data.Calculate the temporal and spatial distance between the track points to find the implicit repetitive data.For those points where space is too small,we use fewer points to replace them,so that the purpose of eliminating duplicate data.The second method is to supplement the incomplete trajectory.First,we need to cut all the trajectories of a class of moving objects into segments by time granularity.Then,we mine the frequent sequences of trajectory points in each segment to fill sparse and incomplete trajectories.Through these two methods,the basic goal of data pretreatment of spatio-temporal trajectory data is achieved,which improves the standardization of data and lays the foundation for further analysis.2.In order to explore the movement law of the moving object deeply,on the basis of the previous research,we should further explore the periodic pattern of the moving object.From the perspective of spatial and temporal changes during the movement of the moving object,we propose a new method to detect the moving object periodic pattern in the trajectory data where the sampling frequency is unstable and the sampling point is sparse in this paper.This method separates the temporal and spatial attributes of the moving object and then processes theyseparately.First,the spatial properties of the trajectory data of the moving object need to be processed.The high density of trajectory points in the spatial region is the area where the moving objects frequently visit.We must find areas where trajectory points are densely distributed.A density-based clustering algorithm is used for pre-processed spatio-temporal trajectory data to extract the high density regions for moving object.These regions are called Hot-regions.Then,we will process the temporal property of the moving object.This study should be based on the discovery of Hot-regions.For each Hot-regions,we need to take the following steps.First,we randomly select a Hot-regions.The timestamps of all points in the Hot-regions will be preserved.These timestamps are treated as a time series.Second,using a certain time granularity,we divide a continuous time segment into discrete points.Third,we improve the existing method to detect potential period.In this way,the visited period of the Hot-regions is found,which realizes the organic combination of temporal and spatial attributes.3.Based on the above theoretical research,we designed and implemented a software system to mine temporal and spatial trajectory data.The main task of the software system is the effective analysis and mining of temporal and spatial trajectory data,and to meet the needs of practical applications.Chinese bird-watching data is used as the data source for this software system.This software system will implement the three algorithms in this paper,namely: SFP(Selection of Feature Point)algorithms,BFCIT(Based-Frequent Completion of Incomplete Trajectory)algorithms and BHMOPD(Based-Hot Moving Objects Period Discovery)algorithms.The system calls Baidu map API to visually display the results of the study.The authenticity and validity of these algorithms in this paper are proved by this system.The experimental results also show that these methods have a good effect on trajectory data where the sampling frequency is unstable and the sampling point is sparse.The software system implements a combination of theory and practice.