An end-to-end UAV-based framework for automated pavement distress detection and condition assessment

Routine pavement inspection remains a time-consuming and labor-intensive task, particularly when conducted over long expressway sections. Traditional manual surveys and vehicle-based systems often lack the efficiency and scalability required for modern intelligent maintenance management. To address these limitations, this study proposes a rapid UAV-based framework for automated pavement distress detection and condition assessment. The framework integrates a YOLO11-based detection model with a tiling–fusion processing pipeline, enabling efficient recognition of six representative distress types from high-resolution aerial images. Quantitative evaluation of surface deterioration is achieved through pixel–real-world conversion based on standard lane width calibration, followed by automated computation of the Damage Rate (DR) and Pavement Condition Index (PCI). Experiments on UAV imagery collected in Guyuan, Ningxia, demonstrate strong performance, achieving mAP@0.5 = 94.5%, MAE(PCI) = 3.84, and r = 0.93 compared with manual evaluation. The proposed framework offers a fast, accurate, and scalable solution for UAV-assisted pavement inspection, providing technical support for intelligent and data-driven road maintenance.