| Rampant cloned vehicle offenses have caused great damage to transportation man-agement as well as public safety and even the world economy.The traditional cloned vehicle detection method needs the support of hardware.Therefore,the deploy costs of these techniques are too high to popularize.The ubiquitous inspection spots that deployed in the city have been collecting moving information on passing vehicles,which opens up a new opportunity for cloned vehicle detection.Existing detection methods using the data collected by the inspection spots usually need to set an appropriate speed thresh-old and identifies a vehicle as the cloned one by judging whether its speed is beyond the given threshold.However,the traffic conditions behave differently across the regions and change over time.Hence,these methods attain lower precision due to that a fixed thresh-old cannot cope well with the changing traffic situation.Furthermore,the simple detection results based on moving behavior abnormality are insufficient to help the authority to pin-point the cloned vehicles.Discerning the behavior patterns of cloned vehicles and even figuring out the motives behind them,makes more sense to solve the crime of cloned vehi-cle.Given this,based on the trajectory data collected by the inspection spots,we propose a Cloned Vehicle Detection Framework,called CVDF.Main contributions of this paper can be summarized as follows:· Trajectory Extraction and Preprocessing The inspection spots deployed in city traffic crossroads have been gathering the information(e.g.,license plate and the time appeared)of the passing vehicles and the trajectory of these cars need to be extracted from collected information first.Besides,there are some noise data caused by the fault or damage of the inspection spot in the extracted trajectories and this paper design a noise filtering algorithm.Finally,the neighbor relationship between the inspection spot pair is mined from the trajectory data.· Cloned Vehicle Detection This paper present a hybrid cloned vehicle detection method by incorporating speed distribution modeling upon historical trajectory data with moving behavior abnormality analysis within local neighbor trajectories.First,the speed distribution of each road is established upon historical trajectory data,based on which the speed threshold can be derived.Then,a sliding window model is used to monitor the behavior of vehicles continuously.In each time window,the speed threshold is used to find abnormal trajectory fragments or identify the moving behavioral outlier within the local neighborhood,and further accumulate the outlier frequentness of each suspicious vehicle.Finally,the vehicle whose abnormal behavior frequentness exceeds the given threshold is identified as the cloned vehicle.· Cloned vehicle behavior pattern mining After cloned vehicle detection,we pro-posed a cloned vehicle behavior pattern mining algorithm(CVBPM)which differen-tiates the traces of various cars with the same plate number and discerns the spatial-temporal behavior patterns of these vehicles.The traces of cars using the same plate number generally mixed,according to the shortest travel time and the matching de-gree,CVBPM discerns the trajectories of various objects then mines the behavior pattern from the trajectory and identifies the vehicles with high-frequency abnormal driving behavior.All in all,this dissertation focused on the cloned vehicle detection problem and a cloned vehicle detection framework has been proposed.Moreover,we conducted the de-tailed analysis around the cloned vehicle detection and cloned vehicle behaivor pattern mining.The experimental results on the real-world data show that CVDF framework has high detection precision and could reveal cloned vehicles behavior effectively. |
