Study On Techniques Of Camera Calibration Of Video-based Traffic Flow Parameters Detection

The main characteristics of city traffic in our country is a hybrid flow, motor vehicles, bicycles and pedestrians are interactional, traffic chaos lead up to traffic jams, accidents and other traffic problems. To resolve the traffic problems, we must accurately grasp the traffic information which is the characteristic parameters of vehicle, bicycle and pedestrian, such as traffic volume, speed and density. For this reason, these parameters must be effectively detected. Currently, the vehicle detectors there have been can detect traffic volume and lane occupancy of the motor vehicles, but they were unable to detect bicycles and other non-motorized vehicles, regardless is the traditional inductive loop, ultrasonic and microwave detectors or video-based vehicle detectors. For resolving the problem of detecting bicycles, our country researched the conductive rubber bicycle detectors and magnetic pressure detector, but they can only acquire the amount of bicycles, they can not acquire speed and direction information, the detectors are instable and impractical. As the video-based detectors can identify and track the vehicle, can provide a wide range of traffic flow characteristics information, and their fast responses, easy installation and maintenance, recur scenes etc., so the research of the video-based detector cause the value of the domestic and international researcher. In the video-based detection system, camera plays an important role, which is a bridge between two-dimensional image and three-dimensional objects. To accurately determine the traffic flow parameters, it is necessary to know transforming relations between two-dimensional image and three-dimensional object, namely camera calibration. In view of this, this paper researches the techniques of camera calibration of the video-based hybrid traffic flow parameters detection, which is suitable for China's hybrid traffic.This paper is completed based on the"Research on detecting technology on the mixed traffic flow characters by image recognition"item, international cooperation in Science and Technology Department of Jilin Province, and"Research on basic program of traffic jam bottlenecks in big city"task"Urban traffic system optimization and control"item, National Basic Research Program of China (973 Program). According to the mixed traffic characteristics in our country and the status in quo of video-based traffic flow detection, aiming at the existingproblems and shortcomings, we have an in-depth study on image preprocessing, corner detection, camera calibration and software system of video-based traffic flow detection. Mainly contains the following contents:First chapter is Introduction, which introduces the research background and significance, present situation and development at home and abroad, as well as the main contents of this paper.Second chapter is detection of calibration reference point. At first, we introduce the image preprocessing technology and its advantages and disadvantages briefly, which affects the result of corner detection, we determine the appropriate image preprocessing technology to use in this paper; Secondly, we analyze the present situation of image reference point detection, then we combine the strongpoint of Harris algorithm and Fo&&r stner orientation algorithm to propose a Harris-based sub-pixel level corner detection algorithm. The experiment results show that the algorithm is simple, reasonable distribution and high positioning accuracy of the detected corners.Third chapter is camera calibration and two-dimensional reconstruction. We firstly introduce four reference frame of camera imaging model and their interrelations; secondly, we analyze the present situation of camera calibration, and then we present a convenient and effective black-box camera calibration method for the traffic scene. The experiment results show that the method is not only suitable for indoor camera calibration, also suitable for use in traffic scene to detect traffic flow parameters, and this method is precise.Fourth chapter is video-based traffic flow detection system realization. In this chapter, we give a software workflow of video-based traffic flow detection system and analyze the deficiencies of existing moving target positioning technology of the video-based traffic flow detection system, and then put forward a video-based traffic flow detection system, which is suitable for China's hybrid traffic. Further, we give the method of obtaining traffic flow parameters by video, and analyze the factors affecting the measurement accuracy.Fifth chapter is Conclusions and Outlook, which summarizes the full text and prospects the work of further research.