The Research Of Freeway Traffic Flow Detection Technology Based On Video Frequency

The detecting system of highway traffic flow based on video is a sort of intelligent traffic detecting system which adopts the technologies for digital image processing and computer vision. In this thesis, the principle of traffic flow detecting based on video is analyzed, the relevant knowledge about digital image processing and computer vision is introduced, and the study is done on camera demarcation, background updating, target identification and traffic flow parameter detection.There are often certain noises existing in the video image for various reasons like surroundings, equipment performance, etc. so it is required to preprocess the input video image. Through comparing different image processing methods, neighborhood averaging method is selected to do image filtering considering the need of real-time processing. Then edge detection is made by means of combining Laplacian with SOBEL operator. After preprocessing, the image can meet the need of further processing well.Camera demarcation is aimed to figure out the corresponding relation between the coordinate system of camera image and the three-dimensional reference coordinate system of object space. Through analysis and comparison of present various demarcation methods, a demarcation method for camera based on orthogonal transform is brought forth. It removes the anamorphic effect of perspective well by transforming the perspective coordinate system to the orthogonal one.Background updating is the basis of traffic flow parameter detection, a background updating method based on the grayscale frequency statistics is used to analyze the video image data collected by fixed camera; meanwhile, the foreground image is captured by means of background difference algorithm.Car detection is the main part of video traffic parameter detection. After investigation and analysis of present various algorithm, a new image matching algorithm based on coherence theory is brought forth to detect the vehicle speed, and a car flow detection algorithm based on virtual detection line is designed on this basis. Good result is obtained when simulating this algorithm by Matlab software.