A systematic review of computer vision-based assessment of concrete properties

Abstract Recent advancements in image processing and convolutional neural networks have revolutionized the assessment of concrete properties. These techniques provide non-destructive and automated solutions to analyze aggregates, air voids, cement phases, and surface defects in concrete and offer significant improvements over traditional manual methods. CNNs have been developed for tasks like segmentation, classification, and quantification of concrete properties under varied conditions. This paper provides a comprehensive review of the latest developments in computer vision applications for evaluating concrete properties. It explores state-of-the-art image processing and CNN architectures being utilized in concrete properties assessment. Then, the application of these techniques to enhance the precision and reliability of concrete analysis is discussed. The review is organized into various categories based on the specific applications and advancements of CNNs in analyzing concrete properties. Challenges such as data variability, feature overlap, and imaging inconsistencies are critically discussed. The potential directions for future research are proposed, integrating advanced multimodal imaging and deep learning frameworks. This review aims to guide researchers and practitioners in advancing the field of concrete property assessment and structural health monitoring of concrete structures.