Improving Bridge Inspection Performance with an Efficient UAV-Based Automated Defect Detection System

Traditional bridge inspections are often characterized by low efficiency, high costs, and notable safety risks. To address these challenges, this paper presents an integrated unmanned aerial vehicle (UAV)-based system for automated defect detection. A key contribution of this work is the development of a holistic, application-driven solution that combines a custom UAV platform with RCO-YOLOv5, a specialized, lightweight deep learning model. RCO-YOLOv5 is methodically optimized from the YOLOv5s framework to achieve a superior balance of accuracy and efficiency for real-world deployment. Architectural enhancements include an efficient RepC2f backbone, an attention-enhanced C3_MSA neck, adaptive Omni-Dimensional Dynamic Convolution (ODConv), and the normalized Wasserstein distance (NWD) loss function. We detail the design and implementation of our UAV system, which streams live video for real-time processing on a ground station. Comprehensive experiments demonstrate that RCO-YOLOv5 achieves a mean average precision ([email protected]) of 91.0%, a 4.0% increase over its baseline, while being 12.8% smaller and 6.7% faster. The successful field testing of the integrated system confirms its potential as an effective tool for modern structural health monitoring.