Research on Eye Detection and Fatigue Early Warning Technologies

Abstract

Utilizing non-contact eye-detection- and image-processing-based fatigue recognition and early warning technology, this study established a human-computer interaction system based on dynamic expression recognition. The degree of fatigue was determined based on the percentage of eyelid closure (PERCLOS). Facial and eye areas in the key frames for facial expression identification were adopted to determine the expression of fatigue. A skin-color technique was used for face detection. A skin model was later established using the hue, saturation, value (HSV) color model, which was then used to detect the skin color of any given face. The improved circle Hough transform algorithm was applied for use in eye detection. Blink rate (pupil region extraction) and PERCLOS were combined in the detection of fatigue.