Abstract: A new method is proposed to locate driver face feature points by chrominance distribution. Based on the face that human lips are red, the luminance of color images acquired by CCD is normalized and the images are projected into an UV space coordinate system composed of color data. Separating the lip region from the projection of images in accordance to red chrominance, the shape of lips is determined by way of combining edge extraction with red pixels, with the state feature points of mouth determined by way of bracketing and approaching. Then, the eye region is preliminarily determined according to face figure/size. Based on the fact that Chinese eyeballs are mostly black, the same method is used to determine and identify the state feature points during eye closing/opening. Experimental results showed that the proposed method is effective to getting rid of the shadow disturbance around lips and eyes, regardless of whether the luminance is even, with an accuracy offered up to higher than 94.3%. The method is thus regarded as real time, accurate and available to be an important criterion for whether a driver is in a state of tiredness.