The Research On Intelligent Traffic Light Control System Based On Machine Vision

With the improvement of people’s economic level,car ownership increases every year,so the increasing pressure exists in urban roads.By analyzing the intersection situation of urban vehicle traffic,there is a focus problem:the traffic lights have not coordinated with vehicles that it can not take full advantage of lane.At the same time,the lane utilization rate is low.Therefore,it is difficult to use accurate mathematical model and the presetting programs in advance to describe.Most of the traditional traffic system is used to control the traffic in different directions through the timing signal lights.When the traffic is much,using such a traffic control mechanism may cause traffic congestion and traffic paralysis.In order to solve the problem of the existing traffic control method,this subject proposes a kind of intelligent control method which can collect real-time traffic information and use the road traffic capacity effectively.This subject from the intelligent traffic control system of the overall design of the program,the main research contents include lane line detection,the extraction and recognition of turning traffic signs at intersection,traffic flow and vehicle speed measuring at green period and vehicle occupation at red period.Specific studies are as follows:(1)Lane line detection,this subject applied in the area of interest to detect lane line,and executed lane line detection according to Hough transform on the former 30 frames of video,and only drawn the designated starting point and the end of the lane line in 31-50 frames of video and the lane line after 50 frames was drawn to the line which resorted the lane line in 31-50 frames of video;(2)The extraction and recognition of turning traffic signs,the subject found traffic signs in a fixed area between the lane lines based on the detection result of the lane line,and the method of inverse perspective transform was used to correct the distortion of the perspective of traffic signs,and on the basis of Hu moment feature and projection feature extraction method,combined with the minimum Euclidean distance classification algorithm compared and analysed the traffic signs recognition results of a variety of feature extraction methods on the traffic signs library established in this study,finally according to the detection results of lane lines and the identification results of turning traffic signs split the different turning lanes;(3)Traffic flow and vehicle speed measuring at green period,traffic flow statistics in the specified lane,using the energy width of the vertical projection of the foreground image of the virtual detection region judged whether there was a car in the detection area,combined with vehicle counting rules to count the traffic flow of specified lane,and according to the experimental results proved the accuracy of algorithm;in all lanes of traffic flow measuring according to the detection results of lane line,the virtual detection area of each lane was determined,then used the method of inverse perspective transformation correcting distortion of the virtual detection area of each lane,and the vehicle counting methods were used before designated lane to calculate the traffic flow of all lanes and then calculated the traffic flow of all different turning lanes.In the statistics of vehicle speed,according to the actual distance and the time difference of the vehicle entering and leaving the detection area,the vehicle speed can be obtained and then calculated the vehicle average speed of all different turning lanes;(4)The calculation of vehicle occupation,firstly introduced the calculation method of vehicle queue length based on projection when the traffic light was green,then introduced the vehicle occupation of calculation method based on vehicle contour area when the traffic light was red,and the method calculated vehicle occupation according to the ratio of each region in different turning lanes of the vehicle contour area and the area of the trapezoid,ultimately judged the current congestion of road on the basis of the ratio of each region in different turning lanes and the threshold value of the region in different turning lanes.Finally,this subject set up the control system of intelligent traffic light based on machine vision with VC++6.0 and OpenCV platform.The green time can be calculated by the statistics of traffic parameters.At the same time,the intelligent traffic control method presented in this subject is simulated in flash and experiments showed the feasibility of this method.