Recognition And Analysis Of Abnormal Driving Behavior Based On Machine Vision

With the development of social economy and the improvement of transportation system,the number of cars and people’s driving demand are constantly improving.However,the incidence of traffic accidents has always been high,which has caused serious harm to people’s lives and property safety.One of the main factors leading to the high incidence of traffic accidents is the abnormal driving behavior during driving.Therefore,how to prevent traffic accidents from the driver’s point of view has become an urgent problem in recent years.Therefore,based on the machine vision method,this paper puts forward an abnormal driving behavior identification method and a dangerous driving classification method,and designs an interactive system to realize real-time early warning of abnormal driving behavior,thus reducing the occurrence of traffic accidents.The main work of this paper is as follows:1.Research the related technologies of deep learning,and determine the research method of abnormal driving behavior identification task.Then the abnormal driving behavior data set is constructed,and the original images are collected from the network public data set,the network crawler and the simulated driving experiment respectively,and the abnormal driving behavior is marked.Some image enhancement methods are used to expand the sample size and improve the richness of the data set.2.Propose an abnormal driving behavior recognition model.The model is improved on the basis of YOLOv5 network,and SPD-Conv block is used to replace the step convolution operation of the original model,which improves the recognition ability of the model for small targets.The Bidirectional Feature Pyramid Network is used to improve the information transmission and feature fusion ability of the model and improve the overall accuracy of the model.Then it is trained and tested on the abnormal driving behavior data set.The experimental results show that the improved model performs better in the detection task and overall detection accuracy of smoking small targets,with an average accuracy of92.4% and an FPS of 60.4,which can meet the task requirements of real-time accurate identification of abnormal driving behavior in practical applications.3.Propose a dangerous driving classification method based on abnormal driving behavior.By establishing the attenuation scoring model,the abnormal driving behavior is quantified reasonably,and then graded according to the quantitative results,so as to improve the effect of early warning of abnormal driving behavior.4.An interactive system is designed.The interactive design based on PYQT5 realizes the real-time display of abnormal driving behavior identification and early warning,and tests the interactive system.The results show that the system can show good results in practical application.