An Approach For Traveling Companions Discovery Based On License Plate Data Stream

With the rapid development of social economy,there are more and more motor vehicles while people travel more frequently.In China,tens of thousands of cameras are currently deployed at major road crosses in most big cities represented by Beijing and Shanghai,which can generate massive traffic data,including ANPR data,GPS data,etc.By mining effective information from these traffic data to solve urban traffic problem represented by traveling companions has been a hot spot in the field of Intelligent Transportation.As one of the focal points and hot problem of studying in the field of modern traffic,the traveling companions have attracted more and more attention.The typical scenario is that criminal gangs commit crimes by vehicles together and the ideal solution need take measure timely to found suspects vehicles,which requirements a high real-time.In recent years the research based on data stream is becoming a trend.Different from the previous research of traveling companions,this paper proposes a new discovery method whose core is parallel PFID algorithm based on license plate data stream.We make use of Spark Streaming to implement the algorithm,through analysing the relationship between the suspect vehicles over a period of time real-timely.By that,we can find the traveling companion shortly,achieve warning effect timely,thus,the public security department can handler it quickly.The main work of this paper is as follows:(1)This paper proposes a discovery method of traveling companions named PFID.According to the definition of traveling companions,the form of ANPR data and the latest data modes and techniques,we adopt association rule mining algorithm.The PFID algorithm based on license plate data stream adopts the idea of Eclat algorithm and mines frequent itemsets.(2)This paper implements the algorithm with the parallel processing framework Spark Streaming on distributed cloud computing environment.We conduct experimental verification and compare from memory usage and response time,the method can found traveling companions effectively and solve the problems such as memory shortage in standalone mode.(3)We develop a prototype system to real-time discovery the traveling companions.In the prototype system,experiment results are presented in some visual manner:within a table,as points on a map,as bars in a chart or graph,and so on.(4)We implement the traveling companions service.In order to make use of the results,the results are provided with restful Web API and the data formats include Text,XML,JSON,etc.