A novel in-situ rapid crack detection approach for reinforced concrete structures using improved active contour contraction image segmentation method

Reinforced concrete (RC) structures are susceptible to cracking. Comprehensive crack information is essential but laborious to acquire. Computer-based visual detection methods, while promising, demand substantial computing resources and labeled training data, often unmet in real-world scenarios. This paper proposes a novel mathematics-based crack detection method. It features: (1) integrated image preprocessing (denoising/super-resolution) to enhance image quality, and (2) a dynamic localized-region contour contraction technique for efficient, accurate, rapid multi-crack segmentation. By leveraging these innovations, the method eliminates need for extensive labeled images and reduces GPU reliance, using mainly CPUs. Tested RC beam surface crack images demonstrate exceptional edge preservation and segmentation accuracy. This mathematics-driven approach overcomes traditional data-intensive limitations, offering practical in-situ potential for reinforced concrete structural health monitoring.