Detecting Drivers' Fatigue in Different Conditions Using Real-Time Non-intrusive System

Driver's fatigue causes fatal road crashes and disrupts transportation systems. Specially in developing countries, drivers take more working hours and drive longer distances with short breaks to gain more money. This paper develops a new real time, low cost, and non-intrusive system that detects the features of the drivers' fatigue. More specifically, the system detects fatigue from the eye closer and yawning. First, the face landmarks are extracted using the histogram of oriented gradients (HOG). Then, the support vector machine (SVM) model classifies the fatigue state from the non-fatigue. The accuracy of the SVM model presented by the area under the curve (AUC) is 95%. The system is evaluated with 10 participants in conditions that can affect the detection of the face. These conditions are different light conditions, gender, age groups, people wearing reading glasses, and males with beard and moustache around their mouth. The results are very promising, and it is 100% accurate.