Research On Key Technologies Of Intelligent Traffic Light Control At Intersection Based On Machine Vision

With the development of society and the rapid increase of vehicles, urban traffic jam is increasingly serious, especially at intersections. This has caused energy waste and environmental pollution and affected people's health. However, the traditional traffic lights control with timing mode cannot adjust time of traffic lights automatically according to the dynamic traffic, and cause low intersection traffic efficiency, even traffic congestion. In this dissertation, the key technology of traffic signal control based on machine vision, and the main work is summarized as follows.1. The vehicle queue length detection algorithm based on connected region is proposed according to the vehicle behaviors at intersections. In the algorithm, the adjacent frame difference is used to detect the head of the queue and the automatic threshold segmentation method is used to detect the tail of the queue, and after that the length of the queue is calculated base on the camera calibration formula and the vehicle queue vertical pixel distance. The results show that this algorithm is capable in detecting the multi-lane vehicle queue lengths timely with strong robustness, high accuracy and low detection error which is less than 8% when the queue length is below 120 meters.2. The traffic flow and speed detection algorithm at intersection is presented based on machine vision. The detection area is set between stop line and crosswalk, and then the area background image is achieved based on Gaussian mixture model. After that, the background subtraction and the projection curve amplitude pulse methods are implemented to calculate vehicle flow in the detection area. The vehicle speed is calculated according to the frame count that the vehicle be driven into and departed out the detection region, the camera frame rate and the detection region width. The results show that the traffic flow detection accuracy is up to 95%, and the measured velocity can reflect the actual vehicle speed.3. The single-stage and two-stage fuzzy control systems based on queue length, traffic flow and vehicle speed are designed. The single-stage fuzzy control system possesses three parameters above simultaneously and resolves traffic lights timing with many control rules. The two-stage fuzzy control system has few control rules. The first level control system makes traffic flow and vehicle speed as inputs and the second level inputs are queue length and traffic state variable which is exported by the first level. The simulation results show that the two-stage fuzzy control system consumes less amount of computation and effectively reduce vehicle stops and the average vehicle delay time.4. The collaborative simulation method is proposed which based on the Matlab and Vissim software according to the defect of traffic control algorithm simulation method. The parameters of the queue length, traffic volume and average speed are obtained by running the intersection model which was build in Vissim software environment. These parameters then are passed to Matlab soft via Excel spreadsheet. Green time is calculated through the designed fuzzy controller and the obtained traffic parameters and returned to Vissim software via Excel, and then switch to next cycle simulation. The simulation process is vivid, intuitive and real.