A method of extracting eye regions to estimate drowsiness level is presented. The proposed method has two main features. A driver's facial image is obtained by CCD (charge coupled device) camera and infrared light emitting diode (LED) installed in the inner mirror, the other, the ridge of the nose, is detected from facial image, followed by extracting eye regions from the detected nose location. The robust method was obtained for an individual difference and a change of the brightness in the vehicle. Experiments have been performed with facial images of 56 drivers in various conditions. As a result, 100% detection rate was obtained for both eyes. Eye regions were exactly extracted even if the driver's face was irradiated by sunlight. The total time for extracting both eyes is about 20 ms by software process. Extraction at the video rate is only possible with the software. (A*)
Abstract