Fatigue Detection Of Motorists Based On Video Image

The method of motorists' fatigue detection can be divided into contact measurement and noncontact measurement.Contact measurement is that the instruments used in experiment can contact the drivers' body and acquire some physical signs which can be used to judge whether the drivers are tired or not.Noncontact measurement is using parameters collected from vehicle itself or the image of driver acquired by camera to decide the fatigue situation of driver.The paper approaches noncontact measurement method based on video detection.Compared with traditional contact measurement,the method require the driver not to touch the instrument,therefore,make them safer and more comfortable.The paper covers such contents:image preprocessing,human face detection,eye detection,extraction of eye contour, how to describe eye contour using mathematical parameter and how to decide the fatigue situation of eyes.Image preprocessing includes the color image gray processing,denoising image and image histogram equalization Its objective is to decrease the information of image and the noise hidden in image such that the image can be processed more easily.Face detection and eye detection adopt Ada Boost classifier.These classifiers constitute a cascade classifier and using them can get the classing results. Haar-like features used to class.Two stage classifiers are used to detect eyes,the first stage detect the face then the second stage classifier detect the eyes according to the position relationship between face and eyes such that the detection rate get higher.Mathematical morphology is used to extract the contour of eye.The geometry description of eye contour adopts ellipse fourier transform.This transform can use fourier coefficients to represent the eye contour.The experiment prove that few coefficients can describe the eye contour accurately.Parameter PERCLOS can reflect the number of blink in every unit time and it can be used to judge the fatigue situation.In the end,the paper simulate drivers' fatigue situation in PC and the result shows as we expected.