Real-Time Driver Fatigue Detection Method Based on Comprehensive Facial Features

In recent years, there have been frequent cases of vehicle accidents caused by fatigued driving, leading to considerable economic losses and a high number of casualties. Accordingly, it has an important social significance for avoiding accident risks to remind tried drivers to take a break promptly. Fatigue driving detection based on facial feature recognition technology has attracted much attention due to its non-invasive, low-cost, and convenient detection advantages. However, the current fatigue driving technology faces the challenge of balancing real-time performance and accuracy in practical applications. Therefore, a fatigue driving detection model based on deep learning is proposed to address this issue. This model includes modules for object detection, head pose estimation, fatigue detection, and distraction detection. First, this paper proposes a facial feature detection algorithm based on YOLOv5 and FSA-Net, which can quickly detects the driver's eye state, mouth state, and head state. Second, in order to better avoid accident risks, the designed system detects driver distraction behaviors during driving process from both cognitive distraction and visual distraction perspectives based on the facial feature detection algorithm. Finally, a comprehensive dangerous driving behavior detection and warning system based on closed-eye detection, yawning detection, 3D head pose estimation, and object detection is designed by integrating multiple fatigue and distraction indicators. The experimental results show that the developed detection and warning system has high detection accuracy, which can provide timely warning when dangerous driving behavior occurs and helps ensure driving safety.