dc.description.abstract | Recently, the issue of driver assistance for safety becomes more
attractive. In this thesis, we propose a computer vision system for monitoring
the driver’s vigilance.
The proposed system consists of seven parts: (1) developing an active
image acquisition equipment, (2) eye detection, (3) eye tracking, (4) face
detection , (5) face orientation estimation, (6) gaze estimation, (7) vigilance
decision.
In order to deal with various ambient light conditions, we utilize an IR
camera equipped with an active IR illuminator to extract several visual cues
such as close/open, eyelid movement, gaze direction, and face direction. A
probabilistic model is developed to measure human fatigue and to determine
fatigue based on the visual cues. At first, we get face images in the same
background and illumination by utilizing Iterative thresholding to find out the
location of brighter pixels. Second, we can obtain the positions of the eyes by
the Connected-component generation. According to the location of the pupil,
we can clip the eye region to be verified by the SVM (support vector machine)
method. then if there are a fixed numbers of image frames succeeded in
detection mode, we can turn the procedure to tracking mode.
In the experiments, the proposed approaches are evaluated by several
different light conditions such at day and night. From the experiment results,
we find that the proposed approach can stably detect or track the eyes in real
time. | en_US |