ABSTRACT
This study aims to develop a system for detecting corrosion on steel bridge surfaces with high accuracy and speed by utilizing visual inspection results alongside an automated system for visualizing corrosion progression. The proposed system consists of five stages: (1) angle adjustment using phase-only correlation to align images from previous and current years, (2) corrosion detection using YOLO, a high-speed object detection algorithm suitable for large datasets, (3) visualization of progression areas through additive mixture of color method by comparing yearly images, (4) quantitative evaluation of corrosion progression at the pixel level using histogram equalization and HSV space conversion, and (5) automated output for comprehensive visualization. The results demonstrate that corrosion can be detected and progression areas visualized using YOLO and additive mixture of color method. Furthermore, progression areas converted to HSV space enable quantitative evaluation of corrosion progression, making this system effective for automated, large-scale inspections using images from inspection vehicles.
