Automatic Identification and Quantification of Dense Microcracks in High-Performance Fiber Reinforced Cementitious Composite with Deep Learning-Based Computer Vision

摘要

High-performance fiber-reinforced cementitious composites (HPFRCC) feature high mechanical strengths, crack resistance, and durability. Under excessive loading, HPFRCC demonstrates unique dense microcracks which are difficult to identify. This study presents a computer vision method for identification and quantification of cracks in HPFRCC based on deep learning for the first time. The presented method seamlessly integrates capabilities of crack detection, localization, and quantification. The number of cracks and the width of each crack in a picture can be automatically determined using the method without human intervention. This study shows that the presented methods achieves an accuracy of 0.986 for crack detection and an accuracy finer than 50 μm (R2 > 0.984) for quantification of crack width, using 200 pictures of HPFRCC and 200 pictures of conventional concrete in the training dataset with incorporation of data augmentation technique. It is envisioned that this method is also applicable to other materials featuring complex cracks.