Development of an Automated Crack Detection System for Port Quay Walls Using Small General-Purpose Drone and Orthophotos

This study proposes a sensor-based infrastructure inspection system for detecting cracks on port quay walls using orthophotos generated from aerial imagery captured by small general-purpose drones. The system incorporates the object detection algorithm YOLOR (You Only Learn One Representation) to accurately identify cracks as narrow as 1 mm, based on visual data acquired through low-cost onboard imaging sensors. To address the challenges of image resolution and scale alignment, two image division techniques—overlapping image tiling and pseudo-altitude slicing—are introduced. The detection system is designed to be low-cost, reproducible, and suitable for direct implementation by local government personnel. Evaluation using real-world inspection data demonstrated that the system achieved detection accuracy comparable to a well-established commercial system. These results confirm the potential of sensor-driven crack detection workflows as a practical alternative for sustainable infrastructure monitoring.