A Study on the advancement of road traffic estimation and greenhouse gas emission estimation using Vision AI technology

The transportation sector has a high proportion of greenhouse gas emissions following the energy and industrial sectors, and greenhouse gas reduction in the transportation sector is a key task that must be addressed to achieve the national carbon neutrality goal. For this, it is essential to calculate detailed CO₂ emissions by region and by road, but most studies currently rely on standardized traffic measurements or speed sensor-based data, so there is a limit to accurately reflecting complex traffic conditions and vehicle characteristics on the actual road.

 

Therefore, this study aims to improve the accuracy of traffic estimation by utilizing artificial intelligence (AI) technology based on real-time road images. In particular, the vehicle type is one of the most important factors in calculating co2 emissions in automatically identifying not only traffic volume but also vehicle types (passenger vehicles, lorries, electric vehicles, etc.) through vehicle detection and tracking technology using CCTV images on the road. Therefore, this study aims to advance a model that is directly used to calculate CO₂ emissions by using vision artificial intelligence models such as yoLO models.

 

By introducing YOLO-based object detection and tracking technology as its core, this study aims to overcome the limitations of existing structured data-based models. It promoted the development of an advanced model that can operate stably in various environmental conditions such as detection and tracking errors, bad weather conditions such as rain, snow, fog, and daytime and nighttime changes, such as urban boulevard with a wide road width and a large number of lanes. Detection and tracking performance were improved by applying the clustering technique based on the model's own parameter optimization and vehicle distribution characteristics, and in the case of failure to detect due to deterioration of image quality, the reliability was increased by directly performing hand-based vehicle coefficients for quantitative accuracy verification. Through this, the co2 emission calculation and comparison verification were conducted through the existing model, and it is expected that comparison with the existing statistical traffic estimation model will be possible.

 

Although the current research remains in the early stage of model development, it is expected that it will be able to develop into a real-time traffic calculation technology optimized for the actual road operation environment through continuous technological advancement in the future. Furthermore, this study is expected to dramatically improve the precision of predicting greenhouse gas emissions through video-based transportation information collection technology and provide a practical and scalable technical foundation in various fields such as smart city construction and eco-friendly transportation policy establishment.