Passenger Surveillance Using Deep Learning in Post-COVID-19 Intelligent Transportation System.

Area of Science:

• Computer Science
• Artificial Intelligence
• Transportation Engineering

Background:

• The COVID-19 pandemic highlighted the need for enhanced public health measures in transportation.
• Existing Intelligent Transport Systems require adaptation to enforce new health protocols like social distancing and mask usage.
• Traffic surveillance systems need advanced capabilities to monitor passenger density and adherence to health guidelines.

Purpose of the Study:

• To propose a deep learning-based framework for real-time surveillance of in-vehicle health protocols.
• To accurately detect the number of passengers and facial mask usage within vehicles using traffic camera feeds.
• To establish a novel method for monitoring in-vehicle social distancing in the post-pandemic era.

Main Methods:

• Development of a deep learning framework utilizing augmented image datasets.
• Application of Transfer Learning techniques with innovative image variations for model training.
• Implementation of fast and accurate detection algorithms for passenger count and mask detection.
• Evaluation using numerical metrics, visual comparisons, and statistical hypothesis testing.

Main Results:

• The proposed framework achieved high testing accuracy ($$) under diverse conditions.
• Demonstrated fast and accurate detection of passenger numbers and mask usage from traffic camera data.
• Achieved $$ accuracy on the Real-Time-Medical-Mask-Detection dataset, showcasing strong generalization capabilities.

Conclusions:

• The developed deep learning framework effectively supports Intelligent Transport Systems in enforcing post-pandemic health regulations.
• This system represents a pioneering approach to monitoring in-vehicle social distancing through advanced image analysis.
• The framework's accuracy and generalization ability offer a robust solution for public health surveillance in transportation.