컴퓨터비전 기반 내화뿜칠 두께 자동 측정 모델 개발

Conventional inspection of fireproofing spray coatings in steel structures relies on manual thickness measurements using gauges, typically based on sampling at a limited number of arbitrarily selected locations. As a result, inspection outcomes may vary depending on the inspector’s experience and judgment. To address these limitations, this study proposes a vision AI–based automated thickness measurement model for the quality inspection of fireproofing spraying work. The proposed model utilizes an RGB-D sensor to generate a three-dimensional point cloud of the work and reference surfaces, and applies Random Sample Consensus (RANSAC) and Singular Value Decomposition (SVD) –based plane estimation algorithms to generate the reference plane. By computing the distance between the reference and work surfaces, the model achieved an average measurement error of 1.5 mm in thickness estimation. Finally, the thickness evaluation criteria specified in KS F 2901 were applied to assess the compliance of each measured point. The proposed model is expected to be applicable to fireproofing spray robots with vision AI-based inspection systems, thereby enhancing the objectivity and consistency of quality assessment.