The Research On Traffic Anomaly Detection Technology In Road Network

Driven by the rapid spread of mobile internet and GPS-embedded smart devices,numbers of trajectories accumulated,these plentiful trajectories senses the traffic dynamic of a whole city.Mining useful information,such as traffic anomaly detection,from trajectories is a current topic in urban computing.Traffic anomalies are caused by accidents,control,protests,sport events,celebrations,disasters and other events.The existing methods detecting traffic anomalies mainly according to the change of traffic flow between city areas or locations,and thus cause low accuracy and not able to detect anomalous road segments timely.In fact,people always travel around the traffic road network.Taking this fact into consideration,this paper studies the problem of traffic anomaly detection in the road network environment and proposes efficient methods.Firstly,this paper proposes a method which is based on the movement feature of trajectories for traffic anomaly detection(TFBTAD).The connectivity of road network result in more than one path between areas or locations,and these paths can be modeled through historical trajectory datasets.Existing works detecting anomalies based on comparing route behavior or other information with historical trajectory database.However,noisy trajectories can be included in historical trajectory data,and thus not able to reveal the normal traffic pattern on road segments.Besides,the change of routing behavior may appears after a long period when anomalous incidents happened.Taking the above shortcomings into account,this paper clusters the trajectories by road segment,and detecting traffic anomalous road segments according to the change degree of vehicle quantity,density and acceleration on road segments.Secondly,this paper proposes another method for anomalous road segment detection,which is based on trajectory flock pattern(FPBTAD).Vehicles travel around road network,and the traffic on road always form trajectory patterns which reveal the travel status of roads.This paper detects trajectory flock pattern,and then computing the feature of flocks such as change of flock quantity and density for anomalous road segment detection.During the preprocess stage of trajectories,this paper also proposes a succinct and rapid map matching algorithm which is based on context of trajectory points(PMatch).After extensive experiments evaluated the performance of these methods on real taxi trajectory data and simulated trajectory data,we come to the conclusion that PMatch algorithm can accurately match the trajectory points on road vertexes.In addition,TFBTAD and FPBTAD algorithm is able to detect anomalous road segment more effectively and timely than existing studies.Besides,TFBTAD have better performance on the aspect of efficiency.In summary,on the basis of analysis of existing researches,this paper makes a lot of theoretical research on the above methods.The research results show that these algorithms have better application prospect on monitor system of urban road traffic.