Research On The Lane Keeping Control System For Intelligent Vehicle Based On Computer Vision

In recent years,due to the growing demand of cars,it leads to some severe consequences,such as damaged environment,traffic congestion,exhaust-emission pollution,etc.Aiming to solve those problems,we ought to do something different in the field of Intelligent Transportation System(Intelligent Transportation System/ITS),which is the key to solve those problem,so a lot of researchers and institutions devote themselves to doing researches in the field of intelligent vehicle.According to the survey,most of the accidents happened due to driver's speeding or reckless driving,therefore,we can change this kind of traditional control mode by using intelligent control theory,which can eliminate human factors,it not only can improve the reliability of the closed loop vehicle control system,but also has significant meaning to improve the traffic stability.Besides,with the booming advanced computer vision technology,we can also apply the computer vision to the intelligent vehicle and take the advantage of it in lane detecting,traffic monitoring and collision avoidance.As we see,the ability to “see” the road is the prerequisite to keep the current lane and cut down on car incidents.The aim of the thesis has two aspects,the one is Lane recognition algorithms and the other is control system of lane keeping based on computer vision.First,we present a novel algorithm to extract the lane from images which come from camera,it is called feature extraction in Region of Interested.First of all,the road image captured by camera,then we use the pre-processing(gray,filtering),contrast enhancement and linear fitting method to complete the reconstruction of the lane line,which can help vehicle to get the current position and locate itself.Vehicle model is divided into four local models based on the speed region,and local fuzzy controllers are built respectively.Then,by using a weighted integration of the four local outputs through a fusion module,a global control output is obtained so as to ensure the stable global performance.To verify the robustness and effectiveness of the proposed multi-model fuzzy controller,simulations of lane keeping are done successfully in Matlab/Simulink,and it is proofed with good performance of keeping the real road.The proposed controller can not only regulate the vehicle fitting in with the road curvature,but also achieve a lateral stability of lane keeping process.As for verifying tests,a micro-car experiment platform is built shown in Chapter 5,in order to verify the robustness and stability of lane detection algorithm.And we also introduces the intelligent vehicle platform based on d SPACE real-time structure,which can pave the way for the future research.