Scholarlink

This study introduces a real-time, non-intrusive driver drowsiness detection system utilizing facial and ocular feature analysis. By leveraging Haar Cascade classifiers with OpenCV algorithms, the system reliably detects eye closure and facial orientation to evaluate driver alertness. The implementation is based on a single board computer, Raspberry Pi 3 platform, processing live video from a PiCam module. This facilitates continuous monitoring and timely alert when fatigue indicators are present. Empirical results indicate robust performance: sensitivity reached 97.87%, specificity was 92.45%, and overall accuracy stood at 95%. The outcomes demonstrate the system’s effectiveness in distinguishing between open and closed eyes states, supporting its viability for embedded automotive applications. The solution’s lightweight, non-intrusive nature addresses practical deployment challenges, particularly where EEG-based methods are unsuitable. By consistently monitoring driver’s behavior and issuing warnings upon detecting drowsiness, the system has significant potential to reduce fatigue related accidents and enhance overall road safety for all road users.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Embedded Lightweight Machine Learning System for Real-Time Driver Drowsiness (Fatigue) Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Thabo Samuel Makhorole,
  • Chunling Tu,
  • Oluwasogo Moses Olaifa

摘要

This study introduces a real-time, non-intrusive driver drowsiness detection system utilizing facial and ocular feature analysis. By leveraging Haar Cascade classifiers with OpenCV algorithms, the system reliably detects eye closure and facial orientation to evaluate driver alertness. The implementation is based on a single board computer, Raspberry Pi 3 platform, processing live video from a PiCam module. This facilitates continuous monitoring and timely alert when fatigue indicators are present. Empirical results indicate robust performance: sensitivity reached 97.87%, specificity was 92.45%, and overall accuracy stood at 95%. The outcomes demonstrate the system’s effectiveness in distinguishing between open and closed eyes states, supporting its viability for embedded automotive applications. The solution’s lightweight, non-intrusive nature addresses practical deployment challenges, particularly where EEG-based methods are unsuitable. By consistently monitoring driver’s behavior and issuing warnings upon detecting drowsiness, the system has significant potential to reduce fatigue related accidents and enhance overall road safety for all road users.