Abstract: A 2-dimensional mathematical model of Gause distribution was developed on the basis of human face complexion to obtain driver's face images using CCD camera and by the clustering analysis in chrominance space. It selects chrominance space as the statistical distribution of complexion and segments face image by maximum intercluster variance threshold to get horizontal and vertical projection of histograms of gradation of face image, thus locating the face. Kalman filter is used to track driver's face by predicting the location of image in next frame in accordance to the face image of driver in present frame and locating it precisely through local search. The experiment results showed that the algorithm to locate and track driver's face on the basis of complexion has high reliability, dynamically locating ability and adaptability to driver's posture.