Empirical Evaluation of Building Crack Identification and Classification using Deep Learning Methodology

Background: Structural health monitoring (SHM) is a critical aspect of maintaining the safety and integrity of infrastructure such as bridges, buildings, and roads. Early detection of cracks and other defects is essential to prevent catastrophic failures and to plan timely maintenance and repairs. This study addresses the challenge of robust crack detection and classification in structural health monitoring systems by leveraging a comprehensive dataset and advanced deep learning techniques. Methods: Utilizing the 'DeepCrack' dataset, which encompasses diverse and heavily annotated images of cracks on surfaces like concrete and asphalt under various lighting and atmospheric conditions, we aimed to develop a model that performs reliably across different scenarios. The proposed hybrid model, CrackNet-Hybrid, integrates Convolutional Neural Networks (CNNs) for feature extraction and Support Vector Machines (SVMs) for classification, combining the strengths of deep learning and traditional machine learning approaches. The dataset underwent extensive preprocessing, including normalization, resizing, Gaussian blurring, edge detection, and segmentation, to enhance the model's performance. Data augmentation techniques further increased the robustness of the model by simulating various real-world conditions. Conclusion: CrackNet-Hybrid demonstrated superior performance, achieving an accuracy of 98%, outperforming other models such as ResNet50, VGG16, and AlexNet.