Deep Learning Model for Driver Behavior Detection in Cyber-Physical System-Based Intelligent Transport Systems

Aim:

    To develop a real-time system for detecting and alerting drowsiness in drivers using YOLOv8 object detection.

Abstract:

      Drowsiness while driving poses a significant risk to road safety, with a substantial percentage of road accidents attributed to driver fatigue. In response to this pressing issue, we developed a driver drowsiness detection system leveraging the YOLOv8 object detection model. The system uses a camera mounted inside the vehicle to classify the driver’s state as either drowsy or awake in real-time. When drowsiness is detected, an alarm is triggered to alert the driver, promoting safety and reducing the risk of fatigue-related accidents. YOLOv8’s speed and accuracy make it an ideal choice for this application, allowing rapid processing and effective classification. Our approach aims to increase driver safety by providing a reliable and efficient system that helps prevent drowsy driving accidents.

Introduction:

       Driver fatigue is a leading cause of road accidents, with drowsy driving accounting for a significant number of crashes and fatalities each year. Fatigue impairs a driver’s ability to stay focused, react quickly, and make sound decisions, creating a dangerous situation on the roads. Traditional methods for detecting driver drowsiness rely on self-assessment or manual observation, which are often unreliable and inconsistent.

       With advances in computer vision and deep learning, automated systems can now monitor drivers for signs of drowsiness. YOLO (You Only Look Once), a real-time object detection algorithm, is well-suited for this task due to its ability to process images quickly and accurately. By mounting a camera inside a vehicle, our system uses YOLOv8 to analyze the driver’s facial expressions, eye movements, and other visual cues that indicate drowsiness.

       This project aims to enhance driver safety by providing a real-time drowsiness detection system. The system is designed to detect signs of fatigue and alert the driver with an audible alarm, reducing the risk of accidents caused by drowsiness. Through this technology, we hope to contribute to safer roads and fewer fatigue-related accidents.

Existing Method:

         Existing methods for detecting drowsiness in drivers often rely on manual observation or self-assessment, which are inherently unreliable. Some systems use biometric sensors or other physiological measurements to detect fatigue, but these methods are often invasive and impractical for everyday use. Early computer vision-based systems typically involved slow and complex algorithms, leading to delays in detection and reduced accuracy.

Disadvantages:

1. Low Accuracy: CNN models often fail to achieve the level of precision needed for reliable drowsiness detection, leading to false positives or missed detections.
2. High Computational Overhead: Traditional CNN-based systems require substantial hardware resources, making them impractical for real-time applications in vehicles.
3. Poor Generalization: Existing systems struggle to adapt to diverse conditions such as varying lighting, camera angles, and individual driver characteristics.
4. Limited Real-Time Capabilities: Many models are not optimized for real-time performance, leading to delays in detection and alert generation.
5. Dataset Dependency: CNN models often rely on preprocessed datasets, which may not adequately represent real-world driving scenarios, reducing their effectiveness.
6. Intrusive Hardware Requirements: Some systems integrate sensors or specialized hardware, which are costly and inconvenient for drivers.
7. Scalability Issues: Existing solutions are difficult to scale for different vehicle types or environmental conditions.

Proposed Method:

         Our proposed method involves the use of the YOLOv8 object detection algorithm to monitor and detect signs of driver drowsiness in real-time. A camera mounted inside the vehicle captures the driver’s facial expressions and eye movements, feeding the data into the YOLOv8 model for analysis. This model is designed to process images quickly, allowing for rapid detection of drowsiness cues.

       When the system detects signs of drowsiness, such as prolonged eye closure or head tilting, it triggers an audible alarm to alert the driver. The use of YOLOv8 provides a robust and efficient solution, capable of real-time object detection with high accuracy. This approach enables a fast response to drowsiness, helping to prevent accidents and enhance driver safety.

Advantages:

1. High detection accuracy for drowsiness indicators such as eye closure and yawning.
2. Real-time performance with low latency, ensuring immediate alerts.
3. Cost-effective implementation using standard cameras and existing hardware.
4. Scalable design suitable for personal and commercial vehicles.
5. Robust performance across diverse environmental conditions, including varying lighting and camera angles.
6. User-friendly interface for easy deployment and operation.
7. Lightweight model architecture, minimizing computational overhead.