Research And Application Of The Algorithm Processing Rain And Fog In Real-Time Video Flow Counting System

Traffic flow counting system has a great significance to forecast the peak period of vehicles and allocate resources reasonably to prevent congestion.Video traffic counting technology is widely used in Intelligent Transportation System(ITS)field because of its low cost,no damage to the road surface and easy deployment and maintenance.However,rain,snow,haze and other extreme weather will cause serious interference to the counting system,so it is very important to study how to ensure the accuracy of video counting in these extreme weather.In this paper,we mainly study the effect of light reflection on rainy road surface and the effect of blur caused by haze on video traffic counting system,we will use effective technical means to improve the accuracy of counting,while ensuring real-time system requirements.Aiming at the interference caused by the reflection of light on the rainy road,this paper presents a threshold-based reflection detection algorithm,which can effectively detect whether the reflective area flicker.In the case of flicker,aiming at the shortcoming of using the detection line method,this paper adopts the regional vehicle tracking method.In the tracking process,this paper focuses on the problems such as vehicle false detection and vehicle adhesion caused by reflective scintillation,which effectively improves the accuracy of system counting.The switching problem between two different tracking processes in different scenarios is also considered.Aiming at the interference of the blurred vision caused by the haze to the counting system,common image de-fog algorithms are studied in this paper.Based on the dark channel de-fog algorithm,a bucket-based atmospheric illumination algorithm is proposed,which can quickly obtain the atmospheric illumination value of the current frame.Aiming at the problem of uneven brightness of image haze-removed,a color enhancement algorithm based on square curve is proposed,which improves the image quality greatly.The accuracy of the counting after removing fog has been improved,but the processing speed of the system can not meet the requirements.In view of this problem,this paper uses CUDA graphics acceleration technology to parallelize the de-fog process,which satisfies the real-time request.