The Algorithm Research Of Vehicle Flow Measuring Based On Video

In the Intelligent Traffic System, parameter measurement of real-time traffic flowis becoming more and more important. Traffic flow parameter includes vehicle flow,vehicle velocity, road-occupied rate and so on, also there are many kinds of ways tomeasure parameters. But the technology of video vehicle detection based on imageprocessing has been a hot field in the Intelligent Traffic System, because it hasprominent advantage that the detection field is extensive and the system setting isflexible. Therefore the detection technology of traffic flow based on video becomes animportant aspect and foundation.Based on reading and understanding the Intelligent Traffic System of national andinternational, the overview of Intelligent Traffic System, the advantages of videomeasurement and development situation of video detection system and algorithm areintroduced firstly. Then the author proposes a novel method of vehicle detection inYUV color model and realizes statistic arithmetic of vehicle flow through studying ofcurrent series of methods and accepting former experience and lesson. During studyingthe algorithm, the following methods are used: image pre-processing, backgroundfoundation and refreshing, color space conversion, shadow detection and so on. In thepaper several common methods of motion object detection are presented and thebackground difference which is used in this paper is introduced in details. For thealgorithm of background difference, background foundation and refreshing is the keyissue, several background updating model is studied and the quick algorithm ofnon-parameter update model is proposed in this paper. In view of shadow, the shadowmodel and detection method are also studied and got rid of shadow in YUV space. Atlast a real-time and reliable traffic flow detection method based on virtualmeasurement thread is designed and realized through programming with Visual C++inWindows operating system. We have tested series of traffic flow image arrays underthe environment of real road, the result is good.