A Proactive Recognition System for Detecting Commercial Vehicle Driver's Distracted Behavior.

Area of Science:

• Computer Science
• Artificial Intelligence
• Transportation Safety

Background:

• Road traffic accidents involving commercial vehicles significantly impact economic development, often linked to driver distraction.
• Current driver distraction surveillance systems have limitations in recognizing diverse objects, scenarios, and behaviors.
• Driver posture is a key indicator of distracted behavior, necessitating advanced monitoring techniques.

Purpose of the Study:

• To develop a comprehensive framework for recognizing driver's distracted behavior by expanding recognition objects, scenarios, and types.
• To analyze driver posture characteristics for improved modeling of distracted behavior.
• To evaluate the performance of Convolutional Neural Network (CNN) models in identifying distracted driving postures.

Main Methods:

• Driver posture characteristics were analyzed to form the basis for modeling.
• Five CNN sub-models were developed for different posture categories.
• A holistic multi-cascaded CNN framework was implemented and tested on extensive daytime and nighttime image datasets.

Main Results:

• Cascaded CNN models (both daytime and nighttime) outperformed non-cascaded models in performance.
• Nighttime models demonstrated faster processing speeds but lower accuracy compared to daytime models.
• The proposed framework significantly enhances the recognition capabilities for driver's distracted behavior.

Conclusions:

• The developed cascaded CNN framework offers a more effective approach to monitoring driver distraction.
• Findings provide valuable insights for designing real-time driver monitoring and warning systems.
• Future work should integrate vehicle state parameters with driver behavior for proactive surveillance systems.