A Real-Time Adaptive Learning Method for Driver Eye Detection

This paper presents an adaptive learning method for real-time driver eye detection, which could be used for monitoring a driverpsilas vigilance level while he/she is operating a vehicle on road. To adapt to the variances in the eye shape and size of different individuals, a learning mode is introduced at the early stage of eye positioning to build the sample learning library. Face detection is firstly performed to narrow the search region. Then contour detection and heuristic rules are used to identify the region of interest (ROI) of the eye. By learning the eye region, a set of images that satisfy the pre-set rules are obtained to form the eye templates. When the number of successful learning exceeds a pre-set threshold, the algorithm switches to the non-learning mode, in which the template matching and a distance factor are used for eye detection. By extracting the skeleton curve and the corner of the eye, the closing level of the eye is calculated which could be used as indicators to the driverpsilas vigilance level. The validation results show that the method could achieve a high level of accuracy.