In a driving support system in ITS, a human interface between a driver and a system is rising in importance. To realize a human interface friendly to drivers, it is essential to detect the driver's state of consciousness or attention. The purpose of this study is to develop the capturing system that capture driver's facial image, and to develop the method of blinking measurement for detecting consciousness degradation, using the motion picture processing.
A robust method against the wide illumination change from daytime to nighttime is essential in a car use. To solve this issue, a weak pulsed infrared light is synchronized to get a CCD camera to capture.
In the blink measurement, the facial and eye areas are extracted from the obtained facial image. After the upper and lower eyelids are detected from the eye areas, blinking is measured from the temporal change of the upper and lower eyelids.
Detection of the upper and lower eyelids is that after the captured image is sliced into vertical sections, candidate points of the upper and lower eyelids are detected in each section. In this method, the candidates of the upper and lower eyelids can be detected reliably irrespective of the shadow of upper eyelid.
We experimented to check the availability of the capturing system in the car experiment. Against the wide illumination change from daytime (2700~32,000[lx]) to nighttime (2~12[lx]), a facial image was taken with a CCD camera when drives at 40km/h. The obtained facial image was verified by visual evaluation and histogram. The experimental results indicate that the obtained facial images are the good images in 30 scenes of which the illumination differs.
In the blink measurement experiment, the extraction rates of this method are 96% for five subjects. Also the extraction rates of this method are over 80% for the subjects wearing sunglasses.