A combined eye states identification method for detection of driver fatigue

For driver fatigue detection system, the correct identification of open and shut eye states matters much to the subsequent calculation of driver fatigue degree. A combined eye state detection algorithm is proposed after a detailed introduction of three algorithms: template matching recognition, upper eyelid curvature calculation, and projection for closed eye detection. First the algorithm of template matching recognition is used as the main method to identify the eye states. Then, before outputting the final result, with an advanced threshold, some images would have a second or third detection for eye states by the use of upper eyelid curvature calculation and projection for closed eye detection. An accuracy of 95.67% for eye states identification in our experiments shows that this algorithm is time saving and robust to illumination and has good prospect of application in a fatigue warning system for drivers.