Recognition And Experiment Of Traffic Congestion State In Rainy Environment

With the development of the economy,urban road congestion is becoming more frequent,and frequent congestion will bring all aspects of social impact.Therefore,the use of scientific methods to study traffic problems has received increasing attention from all society.This thesis focuses on the recognition of traffic congestion in rainy environments.On the one hand,based on the convolutional neural network model and the traffic video information,the road occupancy rate is used as the classification index of the traffic congestion state,and a traffic congestion state recognition model is constructed.On the other hand,considering that some cities or regions are affected by rainy days due to climatic and geographical factors,traffic congestion recognition systems based on video information can be disrupted.Based on the residual neural network,this thesis aims to eliminate the adverse effects of rainy weather environment by learning the characteristic information of rain stripes.An integrated recognition model under rainy conditions is constructed to achieve the purpose of accurately identifying traffic congestion conditions.Finally,an experimental platform for traffic congestion recognition in a rainy environment is established,and the effectiveness is verified through simulation experiments.The main work of the study includes the following three aspects:1.Aiming at the problem of traffic congestion recognition,a recognition model based on multi-scale convolutional neural network is establishedFocusing on the intelligent transportation system,in order to solve the current traffic congestion problem,the pre-exploratory study of traffic congestion recognition is carried out.By introducing the index of road traffic occupancy,this thesis divides traffic congestion into four levels,and then builds a traffic congestion state dataset.A multiscale convolutional neural network is used to design a traffic congestion recognition model.After the model is trained,in the experimental comparison stage,the validity of the proposed model is verified by comparing the traffic congestion recognition methods of related literature.The experimental results show that the model in this thesis can effectively identify the traffic congestion state,and has better performance than other methods.2.Further considering the rainy environment influence,a trafficnet traffic congestion recognition model considering the characteristics of rain stripes is proposedAiming at the problem of traffic congestion state recognition in rainy weather,a trafficnet model was designed to improve the accuracy and robustness of traffic congestion state recognition.The first part of the trafficnet model is used to reduce the impact of the rainy environment on the image and restore a clear image.By studying the characteristics of rain stripes,and then estimating the rain stripes of the rainy day image,combine with the original image to obtain a clear image after restoration.The second part is used to identify traffic congestion.In the experimental comparison phase,the effectiveness of the proposed trafficnet model is verified by comparing the traffic congestion recognition methods in related literature.The experimental results show that the trafficnet model can effectively reduce the interference of rainy environment on traffic congestion state recognition,and has better performance than other methods.3.Design and build a server platform to deploy and validate the effectiveness of the trafficnet modelThis thesis build a lightweight application server under Windows Server 2016 via Flask framework.Based on Android operating system,a client program is designed,which uses HTTP protocol to communicate with the server.The client design respectively involves the basic functions of user's account registration,login,account exit,and account password modification,as well as traffic video upload and server return identification data and other functions.Finally,based on the above platform and the deployed model,the simulation experiment is carried out to identify the traffic congestion state,which verifies the validity of the model.