Design And Implementation Of Fatigue Driving Detection System Based On Facial Features

The development of the modern transportation industry has greatly facilitated the lives of ordinary people,but the huge casualties and losses caused by frequent accidents cannot be ignored.Fatigue driving is one of the main causes of frequent accidents.How to accurately and real-time determine the driver’s fatigue state and timely warn have become a problem that needs to be solved urgently.The purpose of this thesis is to study and design a fatigue driving detection system based on facial features,which can realize fatigue state determination based on face detection algorithm and facial key point location algorithm,and add distracted driving detection function based on smoking behavior.When it is determined that the driver is fatigued or distracted,a voice alarm message will be issued to ensure that the driver is awake and avoid accidents.The main content of this thesis includes facial feature detection research,distracted driving behavior detection research and the implementation and testing of the system.In order to realize the determination of fatigue state based on facial features,the face detection algorithm and key point localization algorithm are studied.Through the MTCNN algorithm combined with the ERT cascaded regression tree model,the positioning of 64 facial feature points is realized,and the fatigue judgment scheme is designed according to the position of the feature points of the eyes and mouth,and the fatigue judgment threshold is set through experiments;through the research on the principles of camera distortion correction and coordinate system conversion,the camera internal parameters are obtained and the conversion from two-dimensional plane coordinate system to three-dimensional space is realized,and then the head in the three-dimensional coordinate system is completed according to the three-dimensional coordinates of the obtained facial key points attitude calculation and threshold setting.In this thesis,the key point positioning model is improved,and 13 new feature points are added at the forehead,which makes the head pose estimation result more accurate.In order to realize the distracted driving detection function based on smoking behavior,the deep learning algorithms of YOLOv5 and YOLOX are deeply studied.Firstly,an improved strategy of adding small target detection layer and input data enhancement is proposed for YOLOv5;secondly,the data set is constructed according to the facial dynamic characteristics displayed in the distracted driving behavior,and then the annotation and expansion of the data set are completed by means of graphic annotation and data enhancement.Finally,the improved YOLOv5 algorithm,the original YOLOv5 algorithm and the YOLOX algorithm were respectively trained in the same data set.After the obtained models were tested and compared,the improved YOLOv5 algorithm with more prominent detection effect was selected to complete the task of distracted driving behavior detection.According to the system requirements and functions,a hardware test platform and software development environment were built,and a fatigue driving detection system based on Jetson Nano was designed;in the system test platform,functional tests and performance analysis were carried out for modules such as fatigue detection and distracted driving behavior detection,the results show that the system can meet the practical application requirements.The fatigue driving detection system based on facial features researched in this paper realizes the function of extracting fatigue features,detecting distracted driving behavior and early warning of drivers during driving,it can pay attention to the driver’s condition in real time while driving and use the form of voice warning to assist the driver to drive safely.