The Research Of Traffic Flow Detection System Based On Image Frequency Spectrum Analysis

With the rapid economic development,the contradiction of " people,vehicles and roads " is more and more serious.This phenomenon not only affected the quality of life of the people,but also caused great distress to the management of society.We can see that the difficulty for the traffic management department also appeared geometric growth.So we must realize the intelligent management of the traffic management system to improve the effici0 ency of the traffic management department.Fortunately,the traffic detection based on machine vision makes it possible for this system to be intelligent.The frequency domain analysis of images is a new image processing technology which developed in recent years.It provides a new perspective for researchers to analyze images,and is widely used in algorithms such as image denoising and recognition.Intelligent traffic is a hotspot in the field of intelligent control in recent years.The detection of traffic data is the key to realize the intelligent control of traffic management system.At present,the mainstream traffic flow detection methods are generally divided into contact detection and non-contact detection.The main method of contact detection is buried coil.And radiation detection and radar microwave is the main method of traditional non-contact detection,but a new method named traffic detection based on video analysis has more advantages.This kind of detection method includes image acquisition,processing and traffic flow analysis,among which the image processing is the most critical.Based on the analysis of a large number of detection methods,this paper proposes a traffic flow monitoring algorithm based on image frequency spectrum analysis.Combining traditional vehicle feature extraction with frequency spectrum analysis,this paper realizes the traffic flow detection.In this paper,the noise reduction,color enhancement,feature extraction and image frequency conversion are improved in image processing.And the contrast test also is improved.Firstly,in the improvement of the image denoising algorithm,the formation and classification of the noise are introduced in detail.A switching function is added to the commonly used median filter function to solve the extreme value and the dispersion characteristic of the salt and pepper noise,which can judge the value of the pixel whether it is a noise and targeted process it.In the improvement of the image enhancement algorithm,the threshold of segmented nodes is determined by flexible adaptive method according to the principle of three-step enhancement,which makes the algorithm better on the monitoring pictures taken at different times.In the image feature extraction,the edge of the foreground object was blurred,which effectively suppresses the energy of the high-frequency energy of the image.In the process of image frequency domain transformation,wavelet transform and discrete cosine transform are combined to improve the proportion of low-frequency energy.Finally,in the improvement of the contrast experiment,according to the near-large feature of the image,a mapping algorithm is used in the vehicle pictures at different locations.Through the final system implementation and experimental testing,the proposed algorithm can effectively distinguish the traffic volume.