Drowsy driver detection system

Abstract

Fatigue is one of the most important reasons for traffic accidents. The drowsy driver detection system (DDDS), steering attention monitor (SAM) etc. were compared. The deficiencies of drivers comfort, resultant veracity and anti-interfere were indicated. The scheme based on infrared, difference image, and Kalman filter was adopted. The pupils' absorptivity of infrared is different at the wavelengths of 850 nm and 950 nm. The interference of other wavelengths light in the background was avoided, and difference images were obtained. With PERCLOS as evaluate standard and the untouched measure, the driver felt comfortable. Kalman tracker assisted by neural net was used to deal with the dynamics of head/eyes movements and made the scheme be real-time. More than 17 frames of effective difference images were obtained in a second. The results showed that the veracity of this scheme to detect driver drowsy is 86.89%.