A Machine Learning Based Driver Monitoring System for the Kayoola EVS

Abstract

With the ever-growing traffic density, the number of road accidents is anticipated to further increase. Finding solutions to reduce road accidents and to improve traffic safety has become a top-priority for Uganda and Kiira Motors Corporation, a state owned automotive company. Therefore, it has become imperative to develop a driver monitoring system which is able to continuously monitor driver behavior. Dangerous driver behavior including distraction and fatigue, has long been recognized as the main contributing factor in traffic accidents. This thesis therefore presents the development of a driver monitoring system which is summarized as follows. Data collection was done by taking pictures of people of different ethnicity, under various lighting conditions, and with different eye states, i.e. open and closed eyes. The dataset was diversified with open source datasets downloaded from the internet. Face detection was done using the Viola Jones Algorithm and a Convolutional Neural Network was designed and trained to classify eye images as open or closed and deviation from a threshold value was determined. Drowsiness detection was then based on how many consecutive frames the drivers eye is closed for, causing a high enough deviation from the threshold value, therefore triggering an alarm.