An Algorithm for Discontinuous Space Vehicle Identification and Tracking Based on Probability Matching and State Prediction
Wan Zhi-pin · Journal of Highway and Transportation Research and Development · 2014
In order to solve the problem of automatic identification and tracking the vehicle target in discontinuous monitoring area of road sections,an algorithm for vehicle identification and tracking in discrete space is proposed.The algorithm is divided into the recognition stage and the tracking stage.At identification stage,it establishes the probability function of a vehicle target in the previous scene by Gauss probability density estimation method.Then,it calculates the matching probabilities of all vehicles in the next scene by probability function,and determines whether there is a vehicle matching the vehicle target of the previous scene by a probability threshold value,and the matching vehicle will be marked and identified in the next scene.At tracking stage,the vehicle is tracked by the method based on unscented Kalman filtering(UKF),and different vehicle targets are tracked simultaneously by state prediction.The algorithm can solve the problem of automatically identify and track the hit-and-run vehicles in different sections under surveillance videos.Experimental tests show that the proposed algorithm has a certain precision on vehicle target identification and can achieve real-time tracking of the vehicle target.