Study On Fatigue Detection Of Motor Driver Based On Face Recognition Technology

With the development of society, motor driver fatigue is increasingly reported as aproblem concerning the road traffic safety. Recent research reveals that approximately 25-30%of vehicle collisions result from driver fatigue. The chance for fatalities is very high.Many ofus have experienced sleepiness while driving. Fatigue may adversely affect drivers'alertnessand ability to drive safe. The study situation on detecting and evaluating techniques of motordriver fatigue in the world (mainly in the U.S.A) is reviewed in this paper.Fatigue driving is a serious problem that leads to thousands of vehicle crashes each year.The drowsiness and fatigue may play an important role in traffic crashes.Of the drowsiness-detection measures and technologies evaluated in this study, themeasure referred to as "PERCLOS"is found to be the most reliable and valid determination ofa drivers alertness level.PERCLOS is the percentage of eyelid closure over the pupil over timeand reflects slow eyelid closures ("droops") rather than blinks.The thesis begins with the coarse detection of face, which is the prework of facerecognition. Through the study of the feature of human skin, we present a method of skindetection. First we convert the image from the space of RGB to C_b'C_r'in order to weaken theeffect of illumination. The Gaussian Mixture Model is set up and maximum likelihoodestimation algorithm is used to estimate its parameters, which is used to judge weather thepixel is a skin pixel or not.According to the distribution of C_bC_r near the eyes, we build the eyesmap at first. With itthe eyes can be detected. The face structure knowledge and gray level distribution informationare use to fine the rest local feature. Then the features are adjusted using a data structure namedFace Bunch Graph. The face is represented by the Gabor jets of the features and their spatialdistances. When driver's sprite states are sleeping, weary and energetic, the shape of eye regionvaries accordingly. On the condition of evident change of the shapes, Elastic Bunch GraphMatching method is selected to recognition the state of eyes, so that we can get the state ofdriver's sprite.