Trajectory Similarity Algorithms And Applications Based On Spatial And Temporal Data

With the rapid development of mobile communication technology,the popularity of portable mobile devices is increasing.Mobile communication devices produce a large amount of data with temporal and spatial information.Pattern analysis and mining of these spatiotemporal trajectory data that can reflect the behavior characteristics of device users has become an important means in the fields of intelligent transportation,smuggling monitoring,public security and intelligent tourism.Trajectory similarity is the basis of trajectory data processing technologies such as trajectory nearest neighbor query,trajectory outlier detection and trajectory clustering,which is of great significance for pattern analysis of trajectory data.Because the sampling quality of different types of trajectory data,such as the sampling time interval and spatial accuracy,is different,the application effect of trajectory similarity algorithm in different quality of trajectory data is different.This paper focuses on the similarity measurement algorithm and its application between low-quality crowd trajectory data and ship trajectory data with different sampling quality.The main work and innovation are as follows:Firstly,a similarity measure algorithm trajectory2vec based on trajectory embedding is proposed.Existing trajectory similarity measurement algorithms are mainly based on the matching of spatiotemporal trajectory points or combined with semantic information to calculate trajectory similarity.However,when the sampling rate and accuracy of trajectory are low,these algorithms often have large errors.This paper proposes a similarity measurement algorithm based on trajectory embedding for crowd trajectory data with low trajectory sampling rate and accuracy.By learning the co-occurrence relationship of trajectory and introducing a priori similarity constraint in the training process,the embedded vector of trajectory is obtained,and the trajectory similarity is measured based on the embedded vector.Finally,the effectiveness of the algorithm is verified by comparing the CDR data with the baseline algorithm.Secondly,STSUM(spatial and temporal similarity with uncertain movement)is proposed.In this paper,for AIS data and radar data with different accuracy and sampling rate,by measuring the motion uncertainty between two points of the trajectory,an ellipse is constructed to represent the possible position of the real trajectory,and then the overall similarity of the two trajectories in terms of matching rate,coverage rate,direction consistency and time similarity is calculated according to the constructed ellipse.Finally,the experiments of trajectory visualization and noise trajectory construction verify that STSUM algorithm is better than baseline algorithm.The algorithm has been applied in practical projects.Thirdly,the adjoint trajectory detection system is designed and implemented.Based on the trajectory similarity measurement algorithm proposed in this paper,a trajectory adjoint detection system is designed and implemented,which includes data storage,data processing,trajectory similarity algorithm model and visual display based on Web.