Zhang, Haixu2024-02-292024-02-292024-02-29http://hdl.handle.net/10393/45989https://doi.org/10.20381/ruor-30191In the last few decades, alkali-silica reaction (ASR) has been extensively discussed in numerous studies as a harmful distress mechanism that poses adverse effects to concrete structures. ASR can lead to the expansion of concrete members, resulting in the formation of cracks on their surfaces, loss in mechanical properties, and severe deformations. These cracks often serve as early indicators of structural degradation, making them crucial for condition assessments aimed at providing preliminary qualitative and quantitative damage extent results. Visual inspection, a commonly used and versatile technique, is employed to assess concrete degradation by visually detecting signs of damage on the surface. However, visual inspection tends to be qualitative and is therefore limited in providing accurate information regarding the extent of damage and its evolution. To address this limitation, researchers have introduced the Cracking Index (CI) as a crack mapping process capable of quantitatively assessing the extent of cracking on the surfaces of affected concrete members. Nevertheless, conventionally collecting data for CI calculation can be both time-consuming and reliant on the expertise of the operators, which can yield inaccurate results when applied in aggressive environments and/or within hard-to-access areas of a structure. To enhance the condition assessment process, the use of Artificial Intelligence (AI) can provide significant assistance by enabling automatic crack detection. However, to improve the performance of crack automating, it is imperative to intensively train a machine learning model, employing a substantial dataset of annotated images, and associated quantitative data. Questions remain regarding the required image quality and image collection methodology to ensure the model’s accuracy and reliability in crack detection. Therefore, this thesis introduces a comprehensive procedure for image acquisition and processing, further analyzing the outputs to explore the capabilities of damage diagnosis through the image-based crack measurement approach.enCondition assessmentCrack automationArtificial intelligence (AI)Alkali-silica reactionConcrete surface crackMachine LearningCracking IndexTotal Crack LengthImage analysisImage processingThesis