Recognition and segmentation of road surface damage using deep learning technology

Tóm tắt

Surface damage detection and segmentation play a crucial role in road maintenance, directly impacting transportation safety and infrastructure durability. This study focuses on leveraging deep learning technology, specifically the YOLOv11n model, to accurately identify and segment surface defects such as cracks and potholes. The YOLOv11n model, known for its lightweight architecture and high-speed performance, was trained on a carefully annotated dataset, employing data augmentation techniques to enhance robustness. Evaluation metrics such as mAP (mean Average Precision), Precision, Recall, and F1-Score were utilized to assess the model's effectiveness. Experimental results demonstrated that YOLOv11n provides high accuracy and efficiency, making it a viable solution for real-time road condition monitoring. The findings underscore the potential of integrating advanced deep learning models into practical road management systems, paving the way for automated infrastructure inspection and proactive maintenance.