Towards Robust Road Quality Detection Using Different Detection Models

Efficient detection of road quality is critical for the safety and longevity of transportation infrastructure. Traditional methods, including visual inspection, and image processing-based filtering techniques, are time-consuming and often fail to accurately capture the complexity of road damage, such as crack shapes and road widths. Our research critically assesses advanced detection models, including YOLOv5, YOLOv6, YOLOv8, and RT-DETR (Real-Time Detection Transformers), focusing on effective road quality evaluation. Our findings reveal that YOLOv8 outperforms its predecessors in detection accuracy, while RT-DETR, leveraging the Vision Transformer (ViT) architecture, offers the highest precision by improving decoding capabilities and optimizing network functions. This study provides essential guidance for selecting the appropriate model based on the trade-off between time efficiency and detection accuracy, thereby enhancing the decision-making process in road quality assessment and contributing to safer transportation infrastructure.