Regresar

Landmark Based Eye Ratio Estimation for Driver Fatigue Detection

Abstract:

In this paper, we present an algorithm for drowsiness detection in drivers of several vehicles based on eye-shape. We use a combination of HOG Linear SVM to locate the face in real-time video, and feature point detection on face region for delimiting the ocular area. The feature point detector use 68-point facial landmark, but we constrain landmarks to the ocular area. We calculate the ocular aspect ratio (EAR), in order to detect driver eye closure, i.e., to detect drowsiness. Results show high sensitivity of the algorithm in the tests performed on a vehicle with a webcam and a warning alert, nevertheless, there is affected by the illumination.