Abstract:PERCLOS value has been used widely in drowsiness detection because of its accuracy and non-contact nature, but in practice, only one PERCLOS criterion is commonly used. In this paper, a method is proposed using continuous eye closure time and PERCLOS value simultaneously for determining the drowsiness degree. Firstly, the algorithm uses Haar-like classifier and Adaboost algorithm for face detection and localization. Then the searching area of the human eyes is narrowed based on human facial structure characteristics. Then the human eyes are positioned using Adaboost algorithm, which can avoid the influence of the eyebrows. Finally image processing methods including image morphology are used to get the vertical height of the human eye, i.e., the distance between the upper and lower eyelids, which can indicate whether the eyes are closing or not. In drowsiness prediction phase, different PERCLOS criteria are used in different time slot. With 10 frames/s testing video speed, the accuracy of the algorithm can reach 86.14%. The method presented in this paper can meet the real-time requirements and improve the accuracy of driver drowsiness degree predictions.