Analysis of a Computer Vision-Based Driver Drowsiness Detection System Under Different Daylight Conditions

This study evaluates a vision-based driver drowsiness detection system under different daylight illumination conditions. A total of 60 real in-vehicle video sequences were analyzed using two drivers under 6 different daylight illumination conditions. For each illumination condition, five 2-minute video sequences were recorded for each driver. The same image-processing procedure was applied to all sequences. Face detection and facial illumination distribution were quantitatively evaluated. The results demonstrate that illumination uniformity should be considered when assessing the reliability of vision-based driver monitoring systems.