Carbon emission prediction and intelligent regulation modeling for traffic systems using vision AI

Article ID: 3441

Keywords: carbon emission reduction; smart traffic signals; machine learning optimization; YOLOv8 object detection; urban traffic management; real-time traffic analysis; environmental impact assessment

Abstract

Urban congestion poses a significant environmental challenge, particularly in developing countries, due to ineffective traffic management leading to elevated vehicle emissions from prolonged idling. This paper introduces a machine learning-based intelligent traffic signal system to cut vehicle idle time and emissions. Using YOLOv8, it detects real-time traffic from four directions. A dynamic phase algorithm replaces fixed signals, counting vehicles and adjusting green lights via capacity ratios. Self-learning, it adapts to scenarios without thresholds. The results showcased notable improvements in traffic efficiency and environmental outcomes, with reductions in delay times ranging from 9% to 65%, particularly a substantial enhancement of 1.94 vehicles per second for the westbound approach. Environmental analysis revealed a 32% reduction in carbon emissions, equivalent to 1502 kg CO₂ saved daily, translating to approximately 548 tons of annual CO₂ reduction. Based on standard carbon market valuations of $15 per metric ton CO₂, this generates 1.5 carbon credits daily with an annual economic value of $8225. The research provides empirical evidence that intelligent traffic management systems can deliver measurable environmental benefits while improving urban mobility. Government authorities validated these findings through official environmental impact assessment procedures. This implementation establishes a scalable framework for smart city initiatives, demonstrating how artificial intelligence applications can address urban environmental challenges through practical, data-driven solutions.