Fatigue Detection Algorithm Based on Eye Multifeature Fusion

Fatigue driving is a critical cause of traffic accidents. Different from a single feature-based fatigue detection algorithm, this article uses the Dlib toolkit to mark facial feature points and presents a novel fatigue detection algorithm considering the eye's multifeature and giving an optimal weight distribution for this fusion model, which intends not only to improve the reliability and error tolerance rate of detection but also to address the issue that traditional algorithms cannot make a balance between real time and accuracy at the same time. The eye's multifeature fusion algorithm simultaneously detects the eye aspect ratio (EAR), the percentage of eyelid closure over the pupil (PERCLOS), the blink frequency (BF), and the pupil occlusion rate (POR). Afterward, the results are normalized to judge the driver's fatigue level and give a corresponding warning. The testing results show that the accuracy of our proposed algorithm reaches 95%, exhibiting a solid potential in fatigue detection applications.

Automated summary

This article proposes a novel fatigue detection algorithm using the Dlib toolkit to mark facial feature points and considers the eye's multifeature. The algorithm achieves an optimal weight distribution for fusion and improves the reliability and error tolerance rate of detection. The testing results demonstrate a high accuracy rate of 95%, indicating strong potential in fatigue detection applications.