International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 06 | Jun 2024
www.irjet.net
p-ISSN: 2395-0072
Real-Time Driver Drowsiness Detection System Based on fast R-CNN Sapana Chandel1, Yogesh kumar Rathore2 1Department of CSE Shri Shankaracharya Institute of Professional Management and Technology Raipur, India 2Department of CSE Shri Shankaracharya Institute of Professional Management and Technology Raipur, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Driver drowsiness is a leading cause of traffic
Network (CNN) with Leaky Fast R-CNN activation functions, utilizing Eye Closure Ratio (ECR) and Eye Aspect Ratio (EAR) as primary indicators. By analyzing eye movements and closures through facial landmarks detected via the face recognition library, effectively monitors driver fatigue. Trained on a dataset of 2000 images, fast R-CNN achieves an impressive accuracy rate of 89.3% in distinguishing between alert and drowsy states. This system enhances road safety by providing immediate alerts to drivers showing signs of fatigue, with future improvements aimed at integrating additional physiological signals and optimizing the CNN architecture for even greater accuracy and responsiveness.
accidents, necessitating the adoption of effective detection technologies. This paper introduces "Fast Region Convolutional Neural Network," a real-time driver drowsiness detection system that uses a Convolutional Neural Network (CNN) with Leaky Faster, yolo R-CNN activation functions, with Eye Closure Ratio (ECR) as the major signal. Fast R-CNN successfully detects driver weariness by analyzing eye movements and closures using facial landmarks identified by the face recognition library. Fast R-CNN achieves an astounding 89.3% accuracy in differentiating between alert and drowsy states. This system improves road safety by providing instant notifications to drivers who exhibit signs of fatigue, with future developments aiming at incorporating additional physiological signals and optimizing the CNN architecture for even better accuracy and responsiveness.
2. LITERATURE REVIEW Driver fatigue warning systems have been integrated into a subset of vehicles due to ADAS's ongoing development and enhancement [4]. Additionally, several automated systems, including Lane Keeping Assist (LKA), Forward CollisionAvoidance Assist (FCA), Intelligent Speed Limit Assist (ISLA), and others, are in operation [5]. These systems enable drivers to temporarily detach from driving responsibilities and assist in managing potentially hazardous situations. In actuality, this assistant's ability to identify hazardous driving situations deteriorates. This is because ADAS depends on sensors that are susceptible to degradation and malfunction in adverse conditions. LKA, which employs sensors to detect the lane lines on the road, may malfunction if the road is inadequately marked. However, while most ADAS systems prioritize environment awareness, minimal effort is devoted to driver monitoring. Although there is anticipation for wholly automated vehicles without accidents to grace the road, technological gaps persist. Moreover, motorist distraction is the primary obstacle to establishing a secure transportation system.
Key Words: Convolutional Neural Network, Eye Closure Ratio, Fast R-CNN
1. INTRODUCTION Worldwide, driver fatigue is a significant factor in traffic accidents, underscoring the necessity for effective and precise detection technologies. In this work, we provide "fast R-CNN," a real-time system for detecting driver sleepiness using an webcam. It uses a Convolutional Neural Network (CNN) architecture as its foundation [1]. By continuously analyzing eye motions and closures, our technology monitors driver weariness and provides prompt detection and alert methods. Many research has examined how machine learning and computer vision detect fatigued driving. Our novel Drowsiness Detection System (DDS) uses OpenCV for realtime video analysis and Keras-based deep learning models trained on several datasets. This device continuously monitors and alerts drivers to weariness [2]. Another model proposed by Wissarut Kongcharoen et al. shows that a CNN with Haar Cascade is the most accurate algorithm (94%) for detecting tired drivers' eye condition and preventing accidents. This Internet of Things-based technology is inexpensive and could improve worldwide road safety [3].
Deep learning techniques, particularly CNNs, have been developed for image-related applications with great success. Deep networks that have achieved success in image classification include AlexNet [6], ResNet [7], and VGGNet [8][9]. However, applying deep learning techniques to signal processing has been relatively sluggish. As a result, in order to leverage the benefits of image-based CNNs (specifically, CNNs) to analyze driver behaviors, we suggest converting the driving signals into multiple images. In order to achieve this objective, we employ the recurrence plot method as a streamlined approach to transform signals into images [10].
Driver drowsiness is a major cause of road accidents, necessitating reliable detection systems. This study introduces "fast R-CNN," a real-time driver drowsiness detection system that employs a Convolutional Neural
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 925