Study On Intelligent Vehicle Lane Detection And Road Searh

With the rapid development of social progress and the road transport system,traffic safety has become an increasingly important social issue. Due to the increasingproblem in current transport system, the Intelligent Transportation System(ITS) hasbeen widely concerned. Intelligent vehicle is a comprehensive intelligent systemwhich integrates the function of environment perception, planning anddecision-making, driving assistance, etc. Intelligent vehicle which represents thedirection toward the development of future vehicle has a highly expansive applicationprospect.In this dissertation, the key technology in the intelligent vehicle environmentperception, lane detection and road detection is studied mainly. It includes image edgedetection, straight line finding, data filtering and tracking, floodfill algorithm, contourfinding and target tracking. Under hardware platform of the electric vehicle of BeijingUniversity of technology BJUT-SHEV, the dissertation establishes a comprehensiveexperimental software platform based on single CCD.First, based on lane detection technology experience of the laboratory, using thevideo stream as experimental material, an algorithm is proposed to screen Houghstraight-line data. Analyze the road information and establish proper screening andtracking modules according to Hough line characters. The algorithm maintains theresent historical results to give the recommended parameters. So the randominterference and detection error is eliminated.Secondly, in order to solve the problem of intelligent vehicle navigation andposition lack of lane guidance information, the road area search technology is studied.Take the advantage of gray characteristics, the algorithm use the floodfill fuction toextract road area first. The edge of the road is often irregular. So a mean-shift targettracking method is proposed. The output is fitted by combining a group of frames. Asa result, stable outputs are remained and bad ones are picked out. Thus, good resultsare achieved in the experiments.Finally, network connection between the vision terminal and the decision-makingand controlling computer is established. According to the model of the cameraperspective feature, the conversion is found between the computer image coordinatesand the vehicle body rectangular coordinates. Convert the acquired data into the datathat decision-making and controlling computer can understand and utilize. Theinformation is exchanged between the terminal and the computer through theself-developed communication terminal and protocol.In order to improve the efficiency of the experiments and provide unifiedexperimental environment, an application is developed, called composite vision experiment platform. This experiment platform provides a variety of debug and testfeatures and the software and hardware interfaces. Every module of the algorithm isinvestigated and validated by a large number of experiments.