Issuu

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395-0056

Volume: 11 Issue: 04 | Apr 2024

p-ISSN: 2395-0072

www.irjet.net

Automated Driver Drowsiness Detection System Using Machine Learning Nagarajan S1, Tamilselvan R2, Veeralokesh B3 1234Dept. of Computer Science and Engineering, Government College of Engineering Srirangam, Tamil Nadu, India

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Ensuring road safety stands as a global priority, particularly given the serious dangers posed by fatigue-related accidents. This study introduces an innovative Real-Time Driver Fatigue Detection System that utilizes computer vision and machine learning techniques. The system integrates a camera to capture live facial imagery, specifically focusing on eye movements for blink rate analysis and mouth actions for yawn detection. Employing Facial Landmark Detection and Convolutional Neural Networks (CNNs), our approach accurately identifies key facial features and computes Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR). Through continuous monitoring and analysis of these ratios, the system promptly detects signs of drowsiness and alerts the driver accordingly. Remarkably, it operates effectively under varying lighting conditions and provides real-time monitoring capabilities while functioning offline for enhanced reliability. Experimental findings confirm the system's effectiveness in mitigating the dangers associated with driver fatigue, thus advancing road safety standards. Key Words: Drowsy Driver Detection System, Facial Landmarks, Driver Safety, alert system, Alert System, Eye Tracking, Blink Rate Analysis. 1. INTRODUCTION Driver fatigue presents a significant global challenge to road safety, highlighted by the United States recording approximately 100,000 incidents annually, leading to 1,500 fatalities and 71,000 injuries. In addressing this issue, drowsiness detection systems have emerged as essential tools. These systems delicately monitor eye and mouth movements, leveraging advanced machine learning algorithms to detect signs of fatigue. By analyzing metrics such as EAR (Eye Aspect Ratio) and MAR (Mouth Aspect Ratio), they prompt drivers to take breaks, thus reducing accident risks. We aim to develop a machine-learning model capable of continuously monitoring these indicators in real time. Utilizing Python, OpenCV, and Keras, this model will discreetly alert drivers upon detecting fatigue, potentially preventing accidents and ensuring the safety of all road users. This proactive approach not only enhances road safety but also fosters a culture of responsibility and care among drivers, contributing to a harmonious and secure transportation environment.

Impact Factor value: 8.226

2. RELATED WORKS This section aims to review the existing techniques for drowsiness detection systems. Vedant Kaushish et al. [1] presented a Driver Drowsiness Detection System that uses OpenCV and Keras to determine if a driver is sleepy based on eyelid movements. The system processes images of eyes under various conditions, totaling around 9,723 images. The dataset is categorized into three parts: frontal face detection, left eye detection, and right eye detection. The model built with OpenCV and Keras showcases high precision and accuracy in detecting driver fatigue, which is crucial for preventing accidents caused by drowsiness. X. G. S. H. X. Zhu et al. [2] authored a paper titled "Real-time Driver Drowsiness Detection Based on Machine Learning Techniques”, which proposes a method for real-time drowsiness detection in drivers using machine learning techniques. The approach likely involves steps such as data collection from various sources like video feeds and EEG signals, and steering wheel movements followed by feature extraction to identify indicators of drowsiness. R. P. Ashish Kumar et al. [3] investigated drowsiness detection using a camera to monitor eye blink rate and eyeball size. Employing CNN-based machine learning for real-time monitoring, it effectively integrates visual cues with advanced techniques. The literature also highlights machine learning algorithms like CNNs for accurately predicting drowsiness levels. S. K. Kushwaha [4] reviews driver drowsiness detection systems, focusing on methodologies like monitoring physiological signals and analyzing driving behavior patterns. The study underscores the importance of these systems in preventing accidents caused by fatigued driving and evaluates various approaches' effectiveness and limitations. B. K. A. J. R. S. Padamata [5] proposes a machine-learning framework for detecting driver drowsiness based on the eye state while driving. The system utilizes the Viola-Jones face detection algorithm to identify the face and extract the eye region from images. A stacked deep convolutional neural network (CNN) extracts features from keyframes and classifies the driver as asleep or awake with a SoftMax layer.

ISO 9001:2008 Certified Journal

Page 2269