Rear Collision Avoidance System using Machine Learning

Abstract

Rear-end collisions pose a significant threat on our roads, often stemming from driver negligence, distractions like texting or eating, and the perilous practice of tailgating. Despite existing safety features in modern vehicles, such as ABS and rear collision systems, these crashes persist, leading to traffic delays, congestion, and potential harm. This initiative addresses this urgent issue through the development of a cost-effective Rear Collision Avoidance System (RCAS). The absence of such a system jeopardizes road user safety and contributes to the financial and environmental consequences of accidents. The proposed RCAS employs a Raspberry Pi 4B, a camera, and the YOLOv8 algorithm for real-time image processing, relying on machine learning algorithms trained on real-world scenarios to identify potential collisions. Customized for the specific challenges of Malaysian roads, the system ensures cost-effectiveness and accessibility, utilizing readily available components like the Raspberry Pi 4 and a Raspberry Pi camera. The core strength lies in providing drivers with real-time alerts upon detecting a potential crash, enabling swift and conscious responses through the immediate activation of an audible alarm. This proactive approach aims to prevent or mitigate the impacts of imminent collisions. This project advances intelligent traffic systems by offering a practical and resource-efficient solution to rear-end collision prevention, ultimately promoting road safety for all drivers. The primary objective centers on developing a cost-effective system to enhance road safety by successfully designing a machine-learning model for early collision detection. The achieved average precision of 80.9% and correct classification of models at 85.5% underscores the system's high accuracy and reliability. Further enhancing real-time responsiveness, the integration of an ultrasonic sensor triggers alerts, such as a buzzer, when objects approach within 50cm. Combining machine learning and sensor technology, this comprehensive approach positions the project as a promising solution for improving overall road safety. The successful synergy of these components highlights the project's potential to revolutionize road safety measures and minimize the impact of rear-end collisions on our streets.