Analysis Of Vehicle Track Prediction Method In Lane Departure Warning System

With the continuous development of the Intelligent Transportation System (ITS) , lane departure warning has become one of the focus in the field. Lane departure warning system include two main parts ,one is the perception of the state of roads and vehicles, the other is the evaluation algorithm of lane departure. This paper make vehicle departure warning system which based on image processing as the research object and with the research of domestic and foreign situation is analyzed, this paper studies all aspects of the experiment and simulation which based on the predicted of vehicle trajectory.Firstly, begin with the modeling building, this paper introduces the Advantage of mathematical simulation software Matlab/Simulink in simulation field .The mathematical calibration of the experimental vehicles and experimental environment include the calibration of image acquisition CCD, the setting of road deviation, and the collection of physical parameters of the vehicle. Based on the kinematics of vehicle, this paper build mathematical modeling for the front-wheel turning angle of vehicle and keep correcting the mathematical modeling which based on kinematics by following the dynamics of vehicle system.Secondly, this paper use kalman filtering theory to observe the state of the system .Then this paper use linear recursive filtering method which is based on finite time data and Systematic mathematic model to recommended that we can use a previous estimates and recent measurements to obtain the unbiased and minimum variance estimation of the current state of the system and reduce the influence of speed for system by using the method of weight distribution of double observer and superposition calculation. By means of the subdividing of vehicle mess region, subsystem of different weight region is established, and each subsystem parameters are adjusted. By changing vehicle mess Interval, the system error is controlled in a smaller range. Actual and calculate trajectory are classified by Euclidean Distance which is based on K-means algorithm. After that we can achieve the on-line estimatation of the mess of vehicle and then the system output is guided. This paper make the road curvature and front-wheel turning angle as an input of system observer to improve the scope of application of the method so that it can adopt both the straight and curve lane.Finally, with the purpose of the coincides of vehicle forecast track and actual trajectory, this paper integrate these subsystem models which is established in view of each mess and verify the reliability of system under various conditions with the method of simulation and experiment...