Abstract:
"Surface cracks in concrete structures pose significant challenges to infrastructure integrity and safety, necessitating accurate and efficient classification methods for timely maintenance. This study addresses the pressing need for automated systems capable of detecting and classifying surface cracks in concrete structures. The study identifies the problem of surface crack classification in civil engineering, emphasizing the limitations of manual inspection methods and the importance of automation. Stakeholders, including civil engineers, construction companies, and government agencies, are key beneficiaries of improved crack detection systems.
To address this problem, the study proposes an optimized image processing model integrated with convolutional neural networks (CNNs) for automated crack classification and severity assessment. The methodology includes image preprocessing techniques such as rescaling and augmentation, followed by CNN-based feature extraction. The model is trained on labeled datasets and evaluated using performance metrics like accuracy, precision, recall, and F1-score. Additionally, severity assessment is performed using computer vision techniques like edge detection and skeletonization to evaluate the width and depth of cracks. This approach aims to enhance infrastructure maintenance practices and improve safety standards.
Initial results demonstrate promising outcomes. The confusion matrix indicates that the model correctly classified 370 negative and 406 positive instances, with only minor
misclassifications. The model achieved an accuracy of 97%, precision of 0.97, recall of 0.97,
and an AUC of 0.99. These metrics highlight the potential of the proposed automated crack
classification and severity assessment system to effectively address critical infrastructure
challenges, providing a reliable tool for stakeholders involved in maintaining concrete
structures."
