Smart Alert System for Drowsy Driver using Haar Cascade Classifier and Dlib Facial Landmark

Abstract

The rise in accident rates involving vehicles such as cars and lorries in Malaysia can be attributed, in part, to driving while drowsy. Various researchers have proposed different techniques for detecting drowsiness, with behavioral-based methods gaining popularity due to their non-intrusive nature. This study focuses to develop and evaluate the accuracy of a behavioral based drowsiness detection system by studying the characteristics of drowsy drivers. The research utilizes the Haar cascade classifier algorithm, Eye Aspect Ratio (EAR) algorithm, and Dlib Facial Landmark Algorithm to effectively detect drowsiness and fatigue. By continuously monitoring the EAR values and identifying when they frequently fall below a threshold value (0.23), the system triggers an alarm sound to alert the driver. The analysis conducted showed that this study achieved a higher level of accuracy, indicating that the algorithms used were highly effective in detecting drowsiness and fatigue with almost 100% accuracy in various conditions, including different lighting conditions (day and night). Consequently, this research contributes to the development of an efficient and reliable drowsiness detection system that can potentially mitigate accidents caused by driver impairment.