Research On Abnormal Behavior Detection System Of Intelligent Car Driver Based On Machine Vision

Nowadays,the number of traditional cars and new energy vehicles continues to rise,although the interior and exterior of the car are gradually improving,and the driving assistance functions are gradually improving,becoming more and more intelligent.However,traffic accidents have not been reduced due to the improvement of automobile intelligent systems.In addition to the low reliability of intelligent systems,human factors are still an important cause.The driver’s driving behavior is not standardized,and there is a lack of timely detection and reminder of the driver’s behavior.Based on the current research vacancies,this paper uses the method of machine vision to detect the abnormal behavior of drivers.The abnormal behavior of drivers usually includes: driver fatigue driving,smoking driving,driving on the phone,using mobile phones,etc.The purpose is to detect irregular driving behavior in real time,and develop a driver abnormal behavior detection system,which is convenient for each car to use.The main work of the paper is as follows:(1)Various image preprocessing methods are studied.The research contents include: 1.The histogram equalization algorithm in image enhancement is analyzed;2.The original image is denoised,and the mean filter denoising and Gaussian filter denoising are compared.(2)The face detection algorithm and the face key point detection algorithm are studied.For face detection,the main features of the face are analyzed,including: geometric features,color features,texture features,HOG features and Haar features.Three different face detection algorithms,SVM,Ada Boost and MTCNN,are compared.For face key point detection,three methods are compared: D-Lib68 feature point detection,TCDCN key point detection algorithm and PFLD key point detection.(3)Focus on the detection and recognition of abnormal driver behavior.The common judgment methods are analyzed for fatigue driving detection,and the PERCLOS algorithm is improved.Secondly,the detection accuracy of smoking behavior,driving a phone call,and playing with a mobile phone has been improved.According to the test results and problems of the R-CNN series detection model and the YOLO series detection model,corresponding improvement measures are proposed.In addition,the driver’s abnormal behavior data set required for this paper is also created,and the abnormal behaviors that drivers will have are classified and discussed.provide support.(4)A driver abnormal behavior detection system is designed and developed.This paper designs the operation process of the driver abnormal behavior detection system,and uses the open source tool Py Qt5 to design the interface,uses the open source development environment anaconda3 environment to build the system,and finally realizes the real-time collection and identification analysis of driver abnormal behavior.detection system.