Research On The Method Of Traffic Flow Information Extraction Based On Video Image Processing

In recent years,road traffic has developed more and more rapidly.The traffic situation in modern cities is complex and changeable,and there are still many problems that need to be improved.Traffic flow information is an important basis for road traffic conditions,and its calculation work should not be underestimated.At present,there are many ways to collect traffic flow information.Among them,the image detection method can not only reduce the information omissions of other methods,but also can be more conveniently transmitted to the transportation department,so that the transportation department can grasp the traffic flow status in real time.Aiming at the research on the intelligent extraction method of traffic flow information,in order to realize the high efficiency of urban traffic flow information collection and the informationization and intelligence of information management,and improve the quality of traffic management and service.With the development of video image processing technology,image acquisition has the advantages of low cost and convenient operation.The introduction of video image processing technology into the collection and detection of traffic flow information has good practical value and development prospects.The main innovations and experimental results of the paper are as follows:Firstly,it innovatively adopts Naive Bayesian’ maximum between-class variance method based on attribute weighting to obtain the best threshold to optimize the background difference method,and applies the improved background difference method according to the characteristics of the target’s shape,size,distribution law,etc.The vehicle target extraction is carried out with the symmetrical frame difference method,and then the difference image is expanded according to the characteristics of the image and the median filter is used to eliminate small noises,and the final target extraction result is obtained.Aiming at the comparative analysis of the extraction results of the two types of methods,current experiments show that both types of methods need to be further improved in accuracy and real-time performance.Secondly,use the latest research and summary of the YOLOv5 network detection algorithm,and apply it to the field of target detection,combining it with two different traditional target detection algorithms for detection work,using such innovative combined detection methods,and The test results obtained are qualitatively evaluated with the extraction results of two types of traditional methods and the evaluation indicators are used for comparative analysis.Through the analogy of qualitative indicators of experiments,it can be known that the YOLOv5 network structure selected in the paper is combined with the improved background difference method and the symmetrical frame difference method.The new algorithm detection results have more excellent accuracy and practicability.Based on the above research,the paper uses Python environment and Open CV to convert video images into real-time traffic flow information,and provides data support for the management and coordination of urban roads,and the obtained road information is more real-time.Improve the automation and scientificity of road traffic flow information management,and provide data reference and engineering practical value for the actual construction and application of road traffic flow information calculations by traffic management departments in the future.