Driver Fatigue Detection Based on Machine Vision*

The eye is the key to detect the driver fatigue, this paper proposes an algorithm to improve the accuracy in the complex lighting environment in the family car. First, facial landmarks are detected based on the ERT algorithm. Eye opening and closure states are detected using HOG feature and support vector machine (SVM). Gaussian mixture model (GMM) is established by calculating values of PERCLOS and ECD to achieve the purpose of fatigue detection. It shows that the algorithm has a certain degree of accuracy and robustness.