Automated Detection of high emission Vehicles in Road Traffic Monitoring Systems

Air pollution caused by vehicle emissions is a growing problem in cities, especially where traffic is dense and existing monitoring systems fall short. In this project, we introduce an automated detection method that identifies visibly high emission vehicles from traffic footage using the YOLOv8 object detection model. Instead of testing the air for emissions, we looked for visible smoke coming from vehicles, which can be seen in real time through video. We used the Roboflow Smoke Vehicles Dataset with 1,034 labelled traffic images, combined with locally gathered emissions footage provided by regional authorities. Our model, trained and run on Google Colab’s T4 GPU, achieved a mean Average Precision (mAP) of 77.7% with inference times under 1 millisecond (0.99 ms). From the results, we believe this model could be useful in real traffic monitoring systems that cities already use. Hence providing a less expensive and more accessible option than sophisticated sensing platforms like EDAR and CARES.