Design Of Highway Driving Environment Safety Risk Source Identification And Risk Assessment System Based On Deep Learning

With the continuous increase of highway mileage and road network density in China,while traffic brings convenience to people,safety has become one of the main problems that cannot be ignored.Among the factors that lead to road traffic accidents,more than23 % are related to road traffic environment,among which road conditions and driving environment have become important factors inducing traffic accidents.Therefore,it is of great theoretical and practical significance for the intelligent management and maintenance of highway traffic facilities to design and develop an efficient and accurate automatic identification and risk assessment system for highway traffic environment safety risk sources.This thesis combines the research content of the Key Laboratory of Road Traffic Safety and Public Security “Intelligent Identification and Risk Assessment Method of Highway Driving Environment Safety Risk Sources Based on Deep Learning”(2021ZDSYSKFKT08).In order to realize the intelligent identification and safety evaluation of highway driving environment safety risk sources,this thesis analyzes the influencing factors and vehicle image representation characteristics of traffic safety risks,studies the identification and risk assessment methods of highway driving environment safety risk sources based on vehicle vision and deep learning,and designs the risk source identification and risk assessment system of highway driving environment safety.The specific research contents are as follows:(1)Research on safety risk source identification method of highway driving environment.A multi-label image classification method based on lightweight convolutional neural network is used to identify the safety risk sources of highway driving environment in the image.By improving the activation function of MobileNetV3 output layer,the MobileNetV3 output layer can output multiple risk source category information at the same time,and realize the identification of highway driving environment safety risk source.In order to improve the recognition efficiency,the attention mechanism of MobileNetV3 is improved and the expansion channel is cut to obtain a compact and efficient classification network.The training and verification of the model on the risk source image sample data set show that the proposed method can effectively identify the risk source of highway driving environment safety and has real-time detection effect.(2)Research on risk assessment method of highway driving environment safety.By establishing a risk assessment model of highway driving environment safety based on the extension matter-element evaluation method,the risk level of highway driving environment safety is determined.Firstly,a highway driving environment safety evaluation index system with hierarchical structure is established,and the weight of evaluation index is determined by analytic hierarchy process.At the same time,the risk source is expressed by the risk source word bag model,and the severity of the risk source is estimated by the multi-layer perceptron classification model to determine the current value of the evaluation index.Finally,a risk assessment model based on extension matterelement evaluation method is constructed,and the safety level of highway driving environment is determined by calculating the correlation degree of different risk levels.(3)Design of highway driving environment safety risk source identification and risk assessment system.Using the Tkinter library in python,a risk source identification and risk assessment system for highway driving environment safety is designed and developed.The system can automatically identify and evaluate the risk sources of highway driving environment safety through the risk source identification and evaluation algorithm based on the video stream data collected by the vehicle camera,and the accuracy of the system is verified by real vehicle experiments.