Vol.31 No.2(1996.6)
Research Report

Drowsiness Level Detection by Measuring Blinks Utilizing Image Processing

Kazuhiko Sugiyama, Morimichi Mizuno,
Tomoaki Nakano, Shin Yamamoto


In recent years, numbers of traffic accidents have been caused by drivers' drowsiness. For preventing traffic accidents, such a system as to detect the drowsiness level without putting any sensor on the driver was developed. According to various reports, as the drowsiness level of a driver decreases, the blinking duration becomes longer than his normal arousal level. In this report, thesefore, we propose a highly reliable method of noncontact measurement for the driver's drowsiness level during driving. The our method consists of the following three parts : 1. We have devised an imaging unit which is less affected by ambient light. We provided an infrared light source which lights the driver's face; this light source emits pulses of light synchronizing with the storage time of a TV camera in electronic shutter mode. 2. We have devised a method of measuring the blinking duration which is independent of the driver's facial features. The blinking duration is calculated from the result of eye-opening measurement using simple gray scale processing. 3. We have proposed a method of estimating the drowsiness level from a change in blinking duration. In this method, the regression line for estimation is repeatedly renewed during driving to respond to individual drivers' differences.
@