Lane Departure Assistance System Based On Machine Vision And Hybrid System

Nowadays,vehicle safety driving assistant technology has become the focus of the research in university and enterprise,and the lane departure warning system is one of the main contents.In order to reduce potential traffic accident due to driver's inattention and strengthen the safety of vehicle during running in the highway,this paper recognizes the lane based on machine vision and designs controller to avoid lane departure,which can not only provide lane departure warning to the driver accurately in real time,but also can control the vehicle automatically to return to the central road.A fast lane recognition algorithm based on Edge Projection is proposed.Firstly,the paper detects vanishing point based on voting mechanism in the region of interest(ROI),defines object search area(OSA)and extracts lane feature points using edge detection.Then these points are projected and counted along the gradient direction,acquiring two points on the lane and estimating confidence to recognize lane.Finally,the strip-shaped region is defined and lane tracking is achieved.Lane departure warning model is established based on machine vision,setting a safe lane boundary,and the time to across lane boundary as a trigger to warn.The human-machine shared control based on hybrid system theory is proposed.The hybrid automaton(HA)is used to design human-machine switching strategy based on EPS in order to achieve human-machine shared coordination.To satisfy steady of system,coordination of human-machine shared and passengers' comfort,the paper solves linear matrix inequality(LMI)to optimize controller,which can control the vehicle back to the center of the road when the driver is inattentive.The car test has showed that the algorithm running on the DSP TMS320DM6437 platform has good real-time performance,robustness and accuracy,the effect of lane recognition and tracking is satisfactory.The proposed control method is evaluated in CarSim/Simulink and hardware in the loop based on CarSim/LabVIEW,simulation results showed that the proposed approach is effective and stable.