Implementation of a Driver Drowsiness Prevention System Using Facial Recognition Technology

Accidents such as drowsy driving and cardiac arrest of elderly drivers are very dangerous. To solve this problem, we proposed a system that switches to autonomous driving mode when the driver falls into an abnormal state. We implemented a drowsy driving detection algorithm using a Raspberry Pi and an IP camera and aimed for low cost and high efficiency. The Raspberry Pi is equipped with a 64-bit Linux operating system, 4GB RAM, a 4-core processor, and a GPU, making it suitable for implementing facial recognition technology. The IP camera transmits video with the H.264 codec via RTSP stream and receives it via the Raspberry Pi's wireless LAN. We processed the video using OpenCV in Python 3.8 and the Miniconda virtual environment and implemented facial recognition using the Dlib library. If the driver's eyes are closed by 75% or less for more than 20 frames, it is considered drowsy driving, and a warning is provided. As a result of the system execution, the system receives the video from the IP Camera, marks key points around the eyes in green, detects normal and drowsy states, and saves logs, warnings, and screenshots. This study demonstrated the potential for the development of autonomous driving safety systems by implementing a real-time drowsy driving detection and warning system based on inexpensive hardware. In the future, it is expected that it will be linked with autonomous driving car systems to quickly enter autonomous driving mode when the driver is drowsy or in other emergency situations, thereby contributing to protecting precious lives and property.