Research Of Accompany Cars Based On Streaming Traffic Big Data

With the development of the vehicle networking,vehicles on the road generate massive traffic behavior data.These data usually exist in two forms.One is automatic license plate identification data,which contains the information of the vehicle itself and the information of the vehicle and the road network.The vehicles' own information includes license plate,vehicle color and vehicle type.The interactive information includes the bayonet camera,the name of the bayonet and the latitude and longitude of the bayonet.Another type of vehicle-related data is GPS data,which is generated in real time by related equipment carried by the vehicle itself.This kind of data records the trajectory of the vehicle.Mining vehicle behavior based on these two kinds of traffic data can help relevant departments find abnormal vehicle behavior in time or make afterwards accountability for abnormal events.There are many kinds of abnormal behaviors of vehicles.Among them,accompany vehicles are more obvious.Through the mining of accompany vehicles,we can have a deeper understanding of the driving behavior of vehicles,and at the same time,we can prevent vehicle tracking behavior,which is very helpful to prevent and detect vehicle crime.However,the structure of the traffic network is complex,and the traffic data is huge.How to timely and accurately find the composition of the accompany vehicles is a problem that needs to be solved urgently.In order to solve the related problems,this thesis establishes a dynamic graph calculation model for real-time discovery of accompany vehicles based on the license plate identification data,which starts from two kinds of traffic data.Based on GPS data,the TSWMD algorithm is proposed and the accompany vehicle groups are efficiently mined by trajectory similarity and trajectory clustering.The main research work of this thesis is as follows:1.This thesis divides the role of the bayonet according to the trajectory data.Firstly,we take the bayonet as a node,get the road network structure graph from the vehicle trajectory,use the improved PageRank algorithm to get the two-dimensional attribute of the bayonet,and classify the bayonet role by clustering the bayonet attribute.Secondly,the bayonet role is added to the vehicle dynamic graph as the constraint condition of static separation to effectively reduce the complexity of the graphics calculation process.Finally,the graph computing relationship is effectively established by taking the vehicle as the node and the accompany relationship as the edge.The dynamic graph computing model is established by SparkStreaming and Spark GrpaphX to mine the accompany vehicle group in real time.2.In this thesis,by introducing the idea of context in natural language processing,the license plate information is mapped to high-dimensional space,and the Node2 vec algorithm is used to calculate the similarity of vehicle trajectory.TSWMD algorithm is proposed to obtain the distance matrix of vehicle trajectory,and to mine the group of accompany vehicles by clustering the vehicle trajectory.The experimental results show that the dynamic graph computing model and TSWMD trajectory clustering algorithm established in this thesis are effective for accurate mining of accompany vehicles.The experimental results show that the dynamic graph computing model and TSWMD trajectory clustering algorithm established in this thesis are effective for accurately mining accompany vehicles.