The Research On Detection And Tracking Techniques For Moving Vehicles In Intelligent Transportation System

Effective detection and real-time tracking of moving vehicles in Intelligent Transportation System (ITS) is the premise of behavior analysis and behavior judgment of moving vehicles. It can real-time accurately provide the traffic flow data of monitoring road, preparing for the subsequent specific treatment, such as finishing and calculation the correlation data of traffic flow, early warning sudden accidents, whole planning transportation system, in order to improve the traffic environment, relieve the traffic jam, raise the road efficiency, raise the efficiency of solving problems.The main research aim of detection and tracking for moving vehicles in ITS is how to raise the accuracy of detection and tracking algorithm for moving vehicles, the response speed and the anti-interference ability. This thesis researches the detection and tracking for moving vehicles in ITS, and designs a real-time detection and tracking algorithm for moving vehicles based on summing up and analyzing the existing methods. In this algorithm, based on establishing the traffic field background model, the moving vehicles are picked out by using the background difference method. According to the principle of Kalman filtering, the target vehicle is tracked.In this thesis, a fast estimation and adaptive updating background model is presented. Its basic idea is: at first, the initial background is fast established by using the statistical method; then, through analyzing the value characteristic of background pixel and prospect pixel in traffic field, the histogram of differential image is used to dynamic obtain the threshold value, and the current instantaneous background image is established; based on the above, the updating algorithm, which is a weighted averaging method of the current instantanous background image and the front background image, is used to estimate the current background image. In this thesis, parameters setting of a vehicle tracking model based on Kalman filtering is determined. At first, the vehicle motion system is analyzed, and system state equation and measurement equation are established. Then, based on that process noise vector and measurement noise vector are gaussian white noise, the concrete expressions of system noise array, observation noise array and the intitial value of filtering error covariance matrix are calculated and determined. Finally, the system simulation is carried out by using the parameter values obtained by learning statistics, and the Kalman prediction data are given.The simulation results show that the algorithm in this thesis has characteristics of the simple mathematical model, fast operation, real-time response, well anti-noise, high precision, and it can meet the demands of the detection and tracking for moving vehicles in ITS.