Assessing the Cracking Development of Ultrahigh-Performance Concrete Using Image Processing

Understanding the patterns and propagation of cracks on the surface of ultrahigh-performance concrete (UHPC) is crucial for revealing the mechanisms of force transfer and failure modes. Despite its significance, research on the relationship between crack evolution and load-displacement curves remains limited. This study presents a new automatic crack detection method developed to assess surface cracks in UHPC under dynamic loading conditions. The method integrates crack pattern identification with kinematic measurement, utilizing image processing techniques such as the rolling ball algorithm and local thresholding to effectively detect cracks. To improve accuracy, a connected component-based approach is implemented to filter out noise. Once cracks are identified, key geometric features—such as crack width, total area, and projections—are continuously evaluated over time. Furthermore, the method introduces a metric for converting pixel-based measurements to real-world dimensions, enabling the determination of actual crack sizes. The accuracy and robustness of the approach are validated through comparisons with experimental results from UHPC subjected to uniaxial tension. This research offers a comprehensive analysis of time-dependent crack characteristics, providing valuable insights into the propagation and failure mechanisms of UHPC.