Research, Based On The Traffic Flow Detection Of Video Image Processing Technology

Traffic flow parameters detection is an important research field ofintelligent transportation system. It includes vehicle flow, vehicle velocity,road-occupied and so on, and there are many kinds of ways to measureparameters. In the numbers of detection techniques, the technology of trafficflow detection based on video image processing has been a hot field in the ITSbecause of it's outstanding advantage that large detection area, informative andflexible system setup. There are broad application prospects.Based on reading and understanding the Intelligent Traffic System ofinternal and external,the overview of Intelligent Traffic System and variety oftraffic flow detection technology are introduced firstly in this paper. Thenintroduced the overview of the development of the video detection technologyand its advantages characteristics in detail, and the development status of thevideo detection system and algorithm described in the summary. This paperfocuses on vehicle detection and statistics algorithm using video imageprocessing method through studying of current series of methods and acceptingformer experience and lesson. The main contents include:Firstly, the algorithm theory used in the algorithm process such as videoimage preprocessing, color space conversion and morphological image processare described. Then the classical algorithms of vehicle target detection arediscussed and conducted thorough research. Experimented on these means byreal-time, spatial accuracy and noise immunity evaluate objectively eachalgorithm and draw advantages, disadvantages and the suitable environment ofeach algorithm.Secondly, according to the problems of background subtraction detectionalgorithms in external scene, this paper proposes a motion detection algorithmbased on a modified k-means clustering. Experiments prove this method notonly can update the background in real-time with the changes of the conditions,can complete the task of moving object detection in complex environment, butalso has good adaptability and robustness. Focusing on the disturbance of moving cast shadow in external scene, a semi-supervised learning method ofmulti-feature fusion based on support vector machine is proposed. Experimentsshow that this algorithm can be able to keep intact the information of the targetvehicle and eliminate the moving shadow effectively. And the performance isalso improved than the traditional methods.Finally, the video traffic detection system has been designed by VC6.0andOpenCV, and the algorithms used in this paper almost have been realized on thePC, and it proved to be available.