Indian Journal of Science and Technology

Abstract

The growing number of traffic accident due to driver’s drowsiness has become a serious problem for society. Hence there is a need to address this problem to avoid accidents by alerting the driver so that road safety can be improved. Invasive techniques that assess physiological conditions, like brain waves, heart rythm rate of the driver and vehicle behavior techniques, including speed, turning angle, lateral position are used in driver fatigue monitoring. In this work, a non-invasive technique that monitors the eye state is used to detect the drowsiness of the driver. A novel feature called minimum intensity projection is proposed to detect the eye state of the driver and Support Vector Machine (SVM) is used to classify the eye state as open or closed. After detecting the eye status, drowsiness level is measured by calculating the duration of eye closeness and if the eyes are found to be closed over some consecutive frames then it is concluded that the person is falling asleep or having a state of drowsiness and hence an alarm is raised. Our approach for fatigue detection is non-intrusive which makes use of only the video from a camera mounted in front of the driver. This work yields a overall fatigue recognition accuracy of 97.4%.
Keywords: Eye State Detection, Fatigue Alertness, Minimum Intensity Projection, SVM