Driver's Eyes State Detection Based on Adaboost Algorithm and Image Complexity

摘要

Eyes state detection is a very important issue in the driver's fatigue detection. In this paper, a novel method is proposed to solve the difficulties and shortcomings in the eyes state detection. Gray-scale transformation and median filtering are used to preprocess images. And then, Adaboost algorithm is used to train the cascade strong classifiers based on the characteristics of Haar to extract the face. In order to reduce the complexity of eyes location algorithm, the only upper part of face is sheared and an image complexity-based algorithm is proposed to locate eyes precisely and solve difficulties to detect the eyes when it closed in current methods. In the proposed algorithm image complexity and maximum connected components are used to locate precise positions of the eyes and judge the state of eyes. The experimental results show that the algorithm has improved the eyes state detection accuracy and has high robustness.