Research On Fatigue Detection Method Based On Deep Learning

Fatigue is one of the main causes of serious traffic accidents in our daily lives.According to data from the National Highway Traffic Safety Administration,due to driver fatigue,approximately 100,000 car accidents occur in the United States each year,resulting in an estimated 1,550 deaths,71,000 injuries and approximately $12.5billion in losses.Another report pointed out that the U.S.government and companies spend US$60.4 billion on fatigue-related driving accidents each year,and the property damage,health hazards,time loss,and reduced productivity caused by fatigue also cost consumers US$16.4 billion.Loss.The German Road Safety Commission(DVR)claims that a quarter of highway fatalities are the result of temporary fatigue of drivers.Driver fatigue has caused huge casualties and property losses.For these reasons,researchers recommend the use of a driver risk early warning system that can determine fatigue.The alarm system can wake a drowsy driver or transfer control of the car to an advanced driver assistance system.With the growing interest in intelligent transportation systems,the development of a practical fatigue detection system is a crucial step.Toyota,Ford,Mercedes-Benz and other automobile companies are also adopting automobile safety technology to prevent traffic accidents caused by driver fatigue.This trend is expected to make cars smarter and significantly reduce accidents caused by driver fatigue.Fatigue detection in the driving environment has the characteristics of complex lighting conditions and rich facial expressions.Because the experimental subject is not in a single environment in the laboratory,the driver’s face is blocked by sunglasses and accessories.Therefore,driver fatigue detection in advanced driver assistance systems(ADAS)is more complicated than fatigue detection in a single laboratory environment.In recent years,in order to improve the accuracy of fatigue detection,scholars have proposed various methods to promote the development of fatigue detection.Although ongoing research shows advancements in fatigue detection technology,there are still many challenges to be solved.The fatigue detection method based on vehicle characteristics and the method based on monitoring steering wheel movement realize various latest technologies.Certain technologies in this group focus on accelerating or interrupting time series,lane departures to determine the level of drowsiness.The fatigue detection method based on biometrics focuses on detecting electronic biological signals,such as EEG(electroencephalogram),ECG(electrocardiogram)and EOG(electrocardiogram).However,the fatigue detection techniques in the two aforementioned categories have serious limitations.The former type of technology can only be used under specific driving conditions and is inherently unstable,while the latter is difficult to practically apply because the driver will feel uncomfortable wearing various signal measurement tools on the body.Therefore,driver fatigue detection technology based on computer vision has become more and more popular.Computer vision technology focuses on detecting closed eyes,yawning methods,and overall facial and head expressions.Aiming at key issues such as feature extraction and fatigue classifier design in driver fatigue detection,this paper takes the physical features of human faces as the object to carry out research on driver fatigue detection based on the convolutional neural network method.This paper introduces the fatigue detection method of multi-physical feature fusion to obtain the driver’s rich fatigue feature information;using the texture information of the driver’s face as the traction,establishes a deep learning model for facial physical feature detection fatigue.The method proposed in this paper is trained and tested on the NTHU-DDD dataset and two other modified non-public datasets.The main result of this work is that the accuracy of fatigue detection is higher than that of other methods,including the original method,with an accuracy rate of over 90%.And it has better generalization ability than multi-physical feature fusion detection method.At the same time,we discussed the fatigue detection method based on convolutional neural network to improve the advanced driver assistance system(ADAS)to make it more robust and reliable decision-making.