Fatigue driving is an invisible killer of road traffic safety. A fatigue driving detection system based on deep learning is pro
posed in this paper to ensure the high efficiency and accuracy of detection. First, OpenCV is applied to gray preprocess the collected
image. Second, directional gradient histogram is used to extract feature and the pre-trained Dlib model is applied to calibrate 68 face
feature points. Finally, the improved detected algorithms of blink, yawn and nod are used to calculate the length width ratio of eyes
and mouth and head posture Euler angle, respectively, and compare with its corresponding threshold to determine whether the driver
is in fatigue state and take early warning measures. Experiments show that the proposed system has 97% accuracy, verifying its ef
fectiveness.
2022, Vol. 1 
