Research On Lane Keeping Control Of Intelligent Vehicle Based On Machine Vision

With the increasing number of vehicles and the increasingly serious traffic accident situation,vehicle safety is getting more and more attention.Relevant research shows that driver erroneous operations are the main cause of fatal driving accidents.In order to improve road traffic safety,vehicle lane keeping control system based on machine vision plays an important role in helping drivers reduce erroneous operations,and has huge market prospects.In order to perform lane line detection and lane keeping control,and based on the analysis of the current status of lane line detection and lane keeping control,this paper designs a region of interest(ROI)prediction method considering vehicle motion states,which improves the real-time and accuracy of lane line detection.On this basis,the lane keeping controller is designed based on model predictive control(MPC)method,and the front wheel angle is optimized to realize the lane keeping control.Firstly,the vehicle-road micro traffic system model,based on 2-DOF vehicle model and vehicle vision preview model,is established,which is used in design of MPC lane keeping controller.The vehicle vision preview model is reasonably simplified,and on this baisis,the ROI prediction model is established,which is used in the design of vehicle-road relative position parameter estimator.In addition,the relevant coordinate system of visual navigation vehicle is defined,and the conversion between coordinate systems is completed.Secondly,the Kalman filtering method is used to establish the parameter estimator,and the vehicle-road relative position relationship is estimated based on the vehicle motion states,and the left and right lane line equation in the world coordinate system is calculated based on this.Combined with the theory of optimal preview time,the centural coordinates of the left and right lane lines ROI are determined,and the coordinates of the left and right lane lines ROI are mapped to the pixel coordinate system according to the geometric relationship of the vehicle navigation related coordinate system.The ROI of both lanes are determineed by offseting from the central position upward,downward,leftward and rightward.On the basis of completing image preprocessing,the lane line detectiin is realized based on Hough transform.Based on the model predictive control method,the lane keeping controller is designed.Considering the effect of lane keeping and vehicle running stability,and considering the physical constraints of the actuator,the multi-objective optimization problem of lane keeping control is transformed into solving a constrained quadratic programming problem,and the input sequence of front wheel angle control is optimized,and the first element in the control sequence is input to the vehicle model to complete lane keeping control.Finally,the traffic scenarios and visual navigation intelligent vehicle system are built in PreScan.The Kalman estimator,lane detection algorithm,and MPC lane keeping controller are completed in MATLAB/Simulink.The vehicle dynamics model is provided by CarSim.Based on PreScan,CarSim and MATLAB/Simulink.a joint simulation platform was built to validate the ROI prediction and lane keeping controller.The simulation results show that the ROI prediction method can effectively reduce the areea of ROI with good real-time performance.The MPC-based lane keeping controller can better complete the lane keeping control and ensure the driving stability of the vehicle.