Research On Vision-Based Traffic-flow Detection Algorithm

With the rapid development of modern society and economy,the acceleration of urbanization,the number of vehicles is also increasing rapidly,social problems such as traffic congestion are becoming increasingly serious.The Intelligent Transportatin System(ITS)has begun to be widely studied by various countries in the world,ITS is the solution to the current situation and the effective way of urban traffic problems.In the ITS,obtaining traffic in real time is the core of the vehicle detection technology.Conventional traffic vehicle flow detection methods include ultrasonic detectors,electromagnetic induction coils,infrared detectors,and microwave detectors,and video vehicle detection technology based on digital image processing has a large detection area and flexible parameter setting.It's widely used because of the advantages of convenient installation and maintenance.The main contents of thesis are as follows:(1)Image preprocessing methods of vehicle flow detection: It mainly includes image graying,image denoising,image binarization and image morphological processing such as erosion and dilation.In terms of image denoising,Gaussian noise or salt-and-pepper noise is added to traffic video vehicles,and then processed by means of mean filtering,median filtering,and Gaussian filtering.(2)Vehicle's moving target detection algorithms: The existing several types of video-based vehicle detection methods such as frame difference method,background difference method,optical flow method are analyzed,and a test comparison is made.According to the advantages and disadvantages of various methods,and combined with specific application environment,this thesis proposes a method of moving vehicle detection based on adaptive mean background modeling combined with frame difference method.(3)shadow detection and elimination: this thesis mainly introduces the mathematical model of vehicle shadow,enumerates the influence of shadow on vehicle detection and several shadow elimination methods.The algorithm of removing vehicle motion shadow by Y component in YUV color space is introduced.(4)Finally,a real-time and reliable traffic detection algorithm based on virtual detection coils is designed.Based on this technology,the algorithm is implemented on the operating system platform using Matlab software to verify the accuracy,real-time,and robustness of the algorithm.The traffic flow video is tested and statistical analysis with regard to the accuracy of flow count,loss detecting rate,and false detecting rate.The results are satisfactory.