Development of a computer vision-based system for recognising fatigue of truck drivers.

Fatigue is inextricable from the environment of truck drivers. Despite the advances in fatigue detection systems, fatigue continues to be one of the leading causes of traffic accidents. Introducing a computer vision-based system for fatigue detection could divert that trend. This graduation project re-searched various computer vision techniques and facial-related fatigue parameters to develop such a system. Through the course of this research combination of yawning, head nodding, and OpenFace 2 as software is examined in detail, and finally, a system is proposed to detect the yawning and fatigue state of the driver. The machine learning algorithm used to determine the yawning state of a face given val-ues obtained from Facial Action Units has the accuracy of 0.961, recall of 0.899 and precision of 0.651. A low precision score is partially corrected by introducing various mitigation approaches. Subsequently, a fuzzy logic model based on yawning and head-nodding frequency is proposed as a model to deter-mine driver’s fatigue.