Research On Intelligent Traffic Signal Control System Based On Visual Sensing Technology

With the rapid development of social and economic,at different times the randomness traffic flow of intersection,complexity and uncertainty,and the change of the flow is irregular,it is difficult to establish accurate mathematical model,the statistical method could not cope with the rapid development of the traffic situation.Therefore,according to the change of traffic flow,real-time adaptive control of traffic signals will become extremely strong.In view of this,this article combined with the existing traffic signal control system,from several major traffic flow detection technology,through applicability than video vehicle detector has good applied prospect,on the basis of in-depth analysis of system requirements and function,respectively to study the video detection module and traffic data analysis module design method,and completed the related experiments,the main research conclusion is as follows:First,from video detection,deep learning is used in the detection mechanism and the video detection mechanism of the restricted boltzmann machine network strategy is proposed.With this mechanism,accurate vehicle detection results can be obtained.In practice,the camera may quiver as the camera may not be completely stationary.The traditional frame differential method and background is the invalid,as the image noise is bigger,the robustness of feature point parameter is very sensitive and aiming at these problems,and the PDE algorithm was proposed based on FAST feature points of video motion compensation algorithm,to solve the problem of camera shake and large noise,also can meet the engineering requirement for real time.In order to further improve the vehicle tracking method and the deep learning algorithm,this paper will be based on the movement of the optical flow method combined with based on random forest algorithm,to solve the problem of tracking accuracy is not high,and achieved good results.Secondly,in the inspection may occur when the data is not accurate and vehicle tracking is not accurate,comprehensive analysis module,designed a comprehensive judgment according to the information of the other module returns of analysis.Based on the hidden markov model(HMM),the dynamic analysis of traffic system is improved and the calculation method of relevant traffic parameters is studied.By effectively linking the relevant data with Hisense signal machine,the automatic real-time detection of traffic flow information is realized,and the signal is controlled by the signal machine.Finally,in the experimental environment,completing the intersection operation situation for the test and comparison test of one year,by optimizing the traffic signal parameters such as length,traffic Gao Pingfeng period,signal cycle to verify the method proposed in this paper the feasibility and effectiveness in practical application,and validate the system in a vehicle environment,day and night light intensity in different environment,the wet weather and other environmental effect on the accuracy of the testing.By the experimental results,test results under various environmental deviation is small,single point of adaptive control method in the case of rush hour and traffic saturation effect is not obvious,but in flat peak and the experimental results show good application effect at night,can satisfy the conditions of large-scale popularization and application.