Research On Driver Fatigue Detection Based On Facial Feature

The proliferation of cars brought convenience and comfort,but also brought issues such as traffic accidents,of which fatigue driving is one of the main causes.Therefore,researching fast and efficient fatigue driving detection has great practical significance to improve road traffic safety.To address the low accuracy of single-feature fatigue detection,errors brought by single-trained fatigue expression models,and the influence of external environment on detection results,this paper proposes a fatigue driving detection method combining YOLOv5 and Dlib to judge drivers’ status and give reminders.Firstly,the method based on driver facial features was chosen after comparing different fatigue detection methods.It is non-invasive and easy to operate.Limitations and low accuracy of fatigue detection using single-feature were addressed by using multifeature fusion method to improve detection accuracy and reliability.Secondly,improved target detection network YOLOv5 is used in this study to detect the driver’s facial region.RFB-s and Focal EIo U replaced SPP and IOU loss respectively to enhance feature extraction and reduce easy samples’ effect on total loss.This algorithm’s testing accuracy is improved by 8.6%.Facial keypoints were located using Dlib facial landmark detection model,and corresponding feature values were calculated to determine fatigue status.Before testing,image preprocessing is needed,with median filtering for noise reduction and flexible LUT functions for illumination compensation.Finally,this paper proposed a multi-index comprehensive fatigue detection method to judge the fatigue state of drivers and give corresponding warning prompt,and the reliability of the fatigue detection method was verified by showing images and data.Experiments showed that the method performed stably in various driving scenarios with an accuracy of more than 90% and a frame rate of 23,meeting the actual demands.