Lane Tracking Control Technology Based On Full Convolution Neural Network And Kalman Filter

In the era of deep integration of intelligent control and industry with technology,the diverless control system has become an indispensable technology in the development and production of modern vehicles,simultaneously,the systems' application has been conostantly improving social production efficiency and improving the publics' travel experience.As one of the most basic and core modules of diverless control,lane maintenance control has important research value and significance.In this paper,deep learning is introduced to lane recognition and detection,least square is introduced to fit the detected lane lines,and Kalman filter is used to track and locate the lane lines.Through these methods,vehicle trajectory on the lane is obtained and lane maintenance control will be finished.The main research contents in the paper are as follows:1.The full convolutional neural network algorithm is used to identify the detected lanes.Firstly,the edge detection algorithm for identifying lane lines in typical computer vision is studied,and an edge detection-based method for marking lane dataset is designed.Then,this paper has complete the dataset production and trained the full convolutional neural network model,and finally realized the detection and recognition of the lane through the trained full convolutional neural network model.2.For the problem of false detection and missed detection in lane detection,the lane tracking and location are studied.First,the lane detection is completed by full convolution neural network,and the lane detection data are obtained.Then,the corresponding aerial view is obtained by inverse perspective transformation,and the pixels of left and right lanes are extracted from the aerial view.The curve equation of the lane is fitted by least square method,and the coefficient of the curve equation is solved.Through Kalman filter,the optimal curve equation is obtained by optimum prediction of the curve equation coefficients fitted each time,completing the lane tracking and improving the accuracy of lane detection.The lane center curve equation and its curve slope are solved according to the left and right lane curve equation,and this curve equation is used as the lanemaintained trajectory.3.To verify the feasibility of lane keeping control,the predictive control model is introduced.The center line of lane tracked by Kalman filter is used as the control trajectory of lane-keeping,and the yaw angle required for driving is calculated and the expected speed is given.The simulation model of MPC controller is constructed to verify and analyze the reliability of the proposed control model.Build a real vehicle test platform,complete the model control tests,and realize the real vehicle test and analysis.