Detection of Road Surface Defects Using UAV-Based Image Analysis Algorithms

Maintaining road infrastructure is essential for economic and social stability. Traditional inspection methods are often expensive, time-consuming, and unsafe, particularly in remote areas. UAVs offer an efficient and high-resolution alternative for monitoring road conditions with minimal disruption. However, automatic road damage detection using UAV imagery remains challenging due to small object sizes, complex backgrounds, and variable environmental conditions. To address these issues, this paper proposes YOLOv8-MSCA, an enhanced version of the YOLOv8n model. The core improvement is the CSPSMAM module, replacing the original C2f block to enable better lightweight feature extraction and robust multi-scale modeling. Additionally, a channel attention mechanism and improved feature fusion enhance the model's focus on critical areas while reducing background interference. The CAFPN structure in the neck network strengthens cross-scale feature interaction, and a new loss function, Wise-MPDIoU, improves localization and boundary prediction for small objects. Experiments on diverse UAV datasets demonstrate that YOLOv8-MSCA significantly improves precision, mean average precision, and F1-score compared to the baseline, while reducing model size. These results confirm its effectiveness and practicality for UAV-based road damage detection.