Autonomous vehicle pollution monitoring: An innovative solution for policy and environmental management

Abstract

Vehicle emissions create growing environmental problems that need creative monitoring methods. This study introduces a real-time pollution monitoring system leveraging deep convolutional neural networks to detect and track vehicles using urban surveillance cameras. By integrating Faster-RCNN and YOLO models, the system estimates vehicle speeds and pollutant emissions based on homography transformations for accurate real-world distance measurements. The system’s robustness is demonstrated through trials, where YOLO outperformed Faster-RCNN in speed and efficiency for urban traffic monitoring. The results imply that real-time emissions data may guide policy choices to lower greenhouse gas emissions, allowing actions such as traffic limitations based on emissions, congestion pricing, and best public transit routes. This scalable, cost-effective system provides a new framework for cities to monitor pollution without requiring additional infrastructure investment, making it particularly relevant for resource-constrained urban environments.