A Study on the Artificial Intelligence Model Using Various Pre-processing Techniques for Crack Detection in Buildings

The structural stability of buildings is crucial for ensuring public safety and preventing catastrophic failures. Cracks in building surfaces pose significant risks, making early and accurate detection a vital aspect of maintenance and safety management. This study introduces a novel artificial intelligence (AI) model based on the Swin-Transformer architecture to detect cracks in buildings effectively. To enhance data quality and model performance, various preprocessing techniques, including Gaussian Blurring, Canny Edge Detection, Sobel Filters, and Laplacian Filters, were applied. These techniques improved noise reduction and edge detection, allowing the model to learn complex crack patterns more efficiently. The proposed Swin-Crack model achieves superior performance compared to conventional convolutional neural network (CNN)-based models, demonstrating an approximately 1.4% higher detection accuracy while maintaining computational efficiency. This was further validated by experiments that showed the model ’ s effectiveness in diverse architectural environments. A key feature of the Swin-Crack model is its ability to operate on standard mobile devices, enabling real-time crack detection without the need for high-performance equipment. This practical solution significantly reduces maintenance costs and facilitates continuous monitoring by on-site personnel. Furthermore, this study highlights the potential of integrating advanced preprocessing techniques with cutting-edge AI architectures to improve safety management in the construction industry. Future research will focus on expanding the dataset and incorporating predictive analytics to evaluate crack progression over time, thereby contributing to long-term structural integrity monitoring. This research provides a scalable and efficient approach to ensuring the durability and safety of buildings.