In an urban environment, traffic congestion, automobile emissions, and road safety are significant problems that lead to financial losses and environmental deterioration. Vehicle detection, counting, speed estimation, and pollution monitoring are frequently handled inconsistently by current traffic monitoring systems. This research closes this gap by applying the cutting-edge YOLOv8 deep learning architecture to create an extensive real-time vehicle recognition and counting system. In addition to calculating vehicle speeds and detecting five different types of vehicles and pedestrians, the system also estimates emission rates in real time using traffic data. After evaluation of two YOLOv8 variants (YOLOv8n and YOLOv8s), it was found that YOLOv8s performed better, with 0.936 precision, 0.822 recall, and 0.930 mAP50 for CNG automobiles. With an emission factor ranging from 0.6 to 0.8, real-time pollution monitoring was made possible by calculating vehicle emissions based on both type and speed. In addition, the system has a web application developed with the Flask framework and allows real-time traffic data display, including emission rates, vehicle counts, and speed calculations. The method is effective, as evidenced by the results, where YOLOv8s exceed YOLOv8n in essential metrics, including the miss-classification rate (as low as 0.112) and F1-score (0.875 for CNG). With its unique method of simultaneous vehicle identification, counting, speed estimation, and pollution monitoring, this research could lead to advancements in road safety, traffic management, and emission reduction