Abstract: This paper describes a practical tiered progressive real-time iris positioning method for mental fatigue monitoring, considering the process of face detection, the positioning of eyes and iris segmentation, is presented According to the characteristics of higher real-time performance demanding, with relatively fixed background and possible head rotation, face tilt and hair occlusion in iris localization for. Skin color model is used to rapidly detect face, ellipse fitting is used to perform face pose estimation and tilt correction, variance projection is used to roughly locate eyes, histogram and custom window function is used to adaptively obtain the iris threshold, and the adaptive threshold segmentation is used for accurate iris positioning. Experimental results show that our method can quickly and accurately locate the iris, can effectively handle the problem of face rotation and tilt in large angle and the facial obscured by hair.