Computer Vision-Based Alert System to Detect Fatigue in Vehicle Drivers

Abstract

AbstractThe most prevailing problem around the world is the increasing number of road mishaps. The number of lives lost on road is much more than any other disease and disaster combined every year. The main cause of accidents to occur is improper and inattentive driving. With proper research on driver drowsiness and the behavioral pattern, the driver exhibits we can reduce accidents. This paper aims to implement the non-intrusive approach to detect the fatigue of the driver and warn the driver immediately to prevent the accident. This paper uses dlib and OPENCV libraries to implement the proposed system. This system uses a dlib face landmark detector to identify 68 distinct spots to apply various face forecast techniques. The live video stream is taken from the camera and is decomposed into continuous frames. The dlib library identifies the eye landmarks and is used to calculate the eye aspect ratio (EAR). The alarm will start given the EAR falls below the threshold value set for some consecutive frames. EAR is obtained by calculating the Euclidean distance between measured eye co-ordinates. OpenCV is used as a primary image processing tool. Python language is used as the main coding language. This drowsiness detection mechanism monitors EAR continuously for drowsiness and gives the alarm sounds if EAR falls below the threshold. Our experimental results show that the proposed system works at a good pace works well in real time. The hardware required for the implementation of this paper is a decent camera that can capture at least 15 frames per second. This paper will help with a significant decrease in the number of accidents.KeywordsFacial landmark identifiersFatigue recognitionEye aspect ratioLive frames extractionEAR threshold