An eye fatigue prediction method is proposed based on the eye aspect ratio, the blink frequency, and pupil diameter. The human eye image is collected by a camera, and the image is processed with gray transform, RGB space segmentation, and Dlib key point prediction. The three characteristic values of the eye are calculated for the method. For six subjects in a continuous eye experiment to collect eigenvalue data, correlation analysis of the eigenvalues was 0.80, 0.87, and 0.63, respectively. The average relative error between the subjective value and the objective predicted value of eye fatigue was 11%.