Intersection Signal Based On Edge Computing And Deep Learning

With the rapid development of the automobile industry and the increasing number of cars,the problem of urban traffic congestion urgently needs to be solved.It is particularly important to optimize the traffic operation capacity of road junctions through relevant technology research under the condition of the existing limited traffic network.In view of the above problems,this thesis adopts computer vision technology to detect pedestrians and vehicles at intersections,improved the detection algorithm of pedestrians and vehicles,and realized real-time statistics of road traffic flow based on edge computing platform.The real-time intersection signal control method is studied to realize the real-time intersection signal control of traffic flow detection.The effectiveness and superiority of the proposed control method are verified by simulation.The main work completed in this thesis is as follows:Analyzed the deficiency of the traditional intersection signal control mode,introduces the research status at home and abroad,subject related technology analysis of intersection signal real-time control system structure and intersection signal control key technology,real-time intersection real-time control performance evaluation indexes are put forward.Aimed at the real-time problems of YOLOv4 target detection algorithm in vehicle categories and pedestrian detection tasks,an improved detection algorithm based on lightweight feature extraction network Mobile Net was adopted to complete the construction of YOLOV4-Mobilenetv3 framework and collected pedestrian and vehicle data sets.The improved detection algorithm model is trained and tested based on transfer learning.The test results show that the proposed algorithm improves the accuracy and speed of vehicle classification and pedestrian detection compared with the original algorithm.Analyzed the requirements of the human vehicle flow detection system,completed the hardware selection and construction of the edge computing platform,uses the improved YOLOv4 target detection algorithm model,carried out software design and implementation of each functional requirement step by step,completed the overall test of the system,and realized the region-based pedestrian vehicle counting.The test shows that the human-vehicle flow detection system adopted in this thesis has high accuracy and real-time performance for the statistics of the number of pedestrians and vehicles,which can meet the statistical requirements of human-vehicle flow based on video detection in intelligent traffic.The intersection real-time signal control problem is studied by combining the intersection pedestrian and vehicle counts and road traffic statistics,and the intersection signal fuzzy controller is designed.The timing strategy of intersection signal control is realized based on fuzzy control rules.Using the simulation software VISSIM to experiment,the experimental results show that the crossing road people real-time detection of intersection traffic signal fuzzy control method can improve the efficiency of road traffic,reduce road traffic stops and delay time,to solve the urban road traffic congestion problems,optimizing the intersection traffic efficiency has a certain practical significance.