Concrete Spalling Depth Change Monitoring Using Directional Lighting Images

Abstract

This study investigates the use of directional lighting images to monitor depth changes in concrete spalling. A neural-network-based photometric stereo method (named NFPS) was applied to images of a surface illuminated from varying angles and directions to generate 3D surface reconstructions. A controlled dataset was created by progressively removing material from a wooden block, representative of concrete, using a CNC machine to simulate increasing levels of spalling severity. Directional lighting images were captured at each stage using a mini-ALICS device, producing three samples: D-0 (undamaged), D-6 (6 mm maximum depth), and D-16 (16 mm maximum depth). In captured images under diffused lighting, damage severity was not visually distinguishable; however, directional lighting enhanced image surface texture and shadows, revealing depth differences. The NFPS method successfully reconstructed the 3D topography of each sample, clearly differentiating between sample spalling depths. From the 3D reconstructions, it was visible that sample D-16 had the most severe spalling depth, despite the spalling surface area from plan view being identical to that of sample D-6. This study’s results demonstrate that directional lighting images, combined with photometric stereo, provide an effective means of monitoring concrete spalling depth changes.