| With the development of the new generation of communication technology,such as GPS/Beidou positioning,vehicle networking,5G and so on,the massive traffic spatio-temporal trajectory data has been rapidly accumulated.It contains extremely rich and valuable laws of traffic travel evolution,which can promote the research of road intelligent transportation system and make it possible to alleviate traffic congestion and exhaust pollution.Spatio-temporal trajectory similarity calculation is one of the most critical research contents whether in clustering,classification,pattern recognition and other data mining methods of traffic spatio-temporal trajectory.Therefore,it has become one of the hotspots and difficulties of scientific research at home and abroad about how to fully,quickly and effectively carry out similarity retrieval and mining analysis of traffic trajectory.On this basis,this paper focuses on the similar trajectory retrieval method based on the semantic cascaded extensible lower bound distance of traffic track big data,and the research on travel preference mining based on latent semantic information of it.And then explore the mining of traffic internal behavior patterns and driving behavior motivation.The specific research work is as follows:(1)Vector representation and similarity modeling of traffic trajectory.Traffic trajectory is a kind of spatio-temporal sequence-graph structure data,which is discrete in both time domain and spatial domain.Compared with general spatio-temporal data,it has the characteristics of significant direction dependence,spatial constraint,spatio-temporal asynchrony and nonstationarity of spatio-temporal distribution,which exists several problems,such as the lack of semantics of model representation,the difficulty of obtaining interest patterns and model evolution.Therefore,this paper deeply studies the collection and processing flow,basic properties and spatiotemporal distribution characteristics of traffic track big data,and proposes a trajectory grid vector modeling method,which lays a solid foundation for further mining and analysis of traffic track big data.(2)Similar trajectory retrieval method based on semantic cascade extensible lower bound distance function.In order to reduce the space-time complexity of DTW and improve the retrieval efficiency,this paper creatively designs a scalable lower bound distance function and proposes a head-to-tail parallel strategy to achieve fast retrieval.Furthermore,in the process of similar trajectory retrieval,according to the potential semantic characteristics of traffic track big data,the concept of conjugate semantic trajectory is introduced,and the early discarding strategy and restricted space search strategy are studied emphatically.Finally,a similar trajectory retrieval algorithm with semantic cascade expandable lower bound distance function is formed,namely,the space-time complexity of each lower bound distance is concatenated from low to plateau.This algorithm effectively solves the problems of large amount of computation and time-consuming in the process of massive time series data retrieval,significantly improves the running efficiency,and verifies the reliability and effectiveness of the algorithm on real data sets.(3)Travel preference mining analysis for latent semantic topic distribution of traffic track big data.In this paper,the topic distribution mining model of traffic trajectory latent semantic information is constructed,which provides a new idea for mining and analysis of drivers' travel preference.In order to avoid the loss of local features of drivers during travel,a geographic cycle travel event extraction algorithm based on conjugate semantic trajectory retrieval is proposed.By combining the conjugate semantic trajectory retrieval algorithm with the topic model,the internal features of the driver's whole travel trajectory are completely preserved,and the mining and analysis of vehicle travel patterns and motivational intentions are eventually realized. |
PDF Full Text Request