Research On Detection Algorithm Of Moving Vechicles

The economic conditions of our country is growing fast,people's living standard has improved than before.The requirements for daily transport are getting higher and higher,traffic safety problems can not be ignored.Especially,the accidents that happened in the highway increase every year.Therefore,the problem of detection and ranging in the highway has been regarded as the research focus by many experts and scholars.Highway is chosen as experimental research background in this paper.According to the research background,some methods of detection and ranging are proposed.This paper is arranged as the follows:(1)Firstly,some traditional vehicle detection algorithms are introduced and their advantages and disadvantages are explained respectively.Then the advantages and disadvantages of various types of camera calibration methods,including traditional and new types of ones,are discussed.They are the theoretical foil for subsequent chapters.(2)An improved three-frame-difference method is used to preprocess the original images and determine the road vanishing point to remove the sky area.Sobel edge extraction and Hough linear fitting achieve detection extract lane edges.Horn-Schunck optical flow method that based on Gauss Pyramid principle detects the front vehicles.The vehicles are finally identified by the combination of Haar and AdaBoost classifiers.(3)The imaging principle of camera is introduced and the geometric model of distance measurement is adopted as front vehicle distance measurement.The measured distance is compared with the actual distance measured by the laser range finder.(4)The system software of preceding vehicles detection and distance measurement based on machine vision is developed on Matlab platform.And the Computer Vision System Toolbox based on Matlab is used to complete this system software.The simulation results show that the system can detect front vehicles more accurately and instantaneity than traditional algorithm.