Research On Highway Traffic Detection Algorithm Based On Deep Learning

Video surveillance is one of the main components of the expressway intelligent traffic detection system.It collects real-time traffic flow data to determine the traffic status of the road and intelligently identify the occurrence of traffic accidents,providing valuable decisionmaking information for the operation management of the road and greatly improving Management efficiency.Vehicle automatic identification technology based on video image sequences is the basis for video surveillance to collect traffic flow data and detect traffic incidents.Therefore,it is of great significance to study a vehicle detection algorithm with high accuracy and good real-time performance to improve the reliability of detection.Therefore,this dissertation studies the vehicle recognition algorithm and the road environment discriminant algorithm based on the deep learning,which greatly improves the accuracy of vehicle detection.Firstly,we collect comprehensive highway monitoring image data in different environments to build a complete database,construct a deep learning model for environmental discrimination,and train the model to realize the real-time environment identification of highways as the basic conditions for vehicle detection and traffic event identification.Then,the vehicle targets were labeled and sample preprocessed for samples in different environments.Based on this,vehicle detection algorithms were studied.In the article,a vehicle detection algorithm based on the weather environment recognition using the fast-RCNN model was proposed to improve the accuracy of highway vehicles detection.Then the performance of the vehicle detection algorithm described in this paper is verified by comparing the differences in algorithm detection performance between different environmental dataset models and global dataset models,different network structures,and multiple detection methods.Finally,taking the Guangzhan line freeway as an example,the video detectors distribution are studied,and the deep learning vehicle detection algorithm constructed in this paper is applied to traffic flow parameters and traffic event detection to verify the performance and feasibility of the algorithm.