The Research And Implementation Of Video-based Traffic Flow Detection Technology

With the rapid development of economic and improvement of the urbanization level, the traffic congestion is becoming more and more serious. In order to realize the scientific and modern management of the traffic and to improve road traffic dynamics, intelligent transportation system (ITS), which is to solve the problem, has attracted extensive attention worldwide. Traffic flow detection system is the base and a very important part of the ITS, which can provide information source to traffic information services and transportation control center. Thanks to the highly development and gradual improvement of some important technologies, such as embedded technology, pattern recognition technology, video image processing technology and DSP technology, the importance of the traffic detection system based on video becomes increasingly evident. The system’s core technology is the real-time detection and tracking of moving target in the video sequence of images collected by the camera monitoring area and attraction of traffic parameters. This paper studies the key technologies of moving target detection and tracking used by the system and endeavors to improve the efficiency of the system.This thesis studies video-based traffic detection technology, introduces the traffic information acquisition and image processing theory and technology, and details the key technologies used by the system flow detection module. The main innovative results include the following four aspects:(1)A background modeling method used an improved k-means clustering algorithm is presented, which can solve the traditional algorithm’s problem to easily fall into local optimal solution and improve the adaptability of the background model;(2)An algorithm based on combination of frame difference method and background subtraction is proposed. Background subtraction is sensitive to light changes, while frame difference method is sensitive to vehicle speed. The improved algorithm effectively improves the accuracy of vehicle detection;(3)A Mean-shift adaptive tracking algorithm based on linear prediction is put forward. The traditional algorithm has too much times of iteration and is more difficult to achieve multiple objectives simultaneously tracking. The improved method reduces the number of iterations, and can simultaneously track several moving targets.(4)A simple traffic statistical system is realized, which can extract the parameters of traffic flow, speed and other traffic accurately;Finally, using the VS2008platform on Windows XP and coding in C++combined with the OpenCV library functions, a video-based traffic detection system has developed. The test results shows the system improves the real-time and accuracy of vehicle detection through the test results.