Research on Building Crack Detection and Extraction Methods Based on Deep Learning and Image Processing Techniques

In this paper, we adopt the U-Net model in deep learning for wall crack recognition and detection, which performs exceptionally well in this field. The U-Net model utilizes its unique encoding-decoding architecture with jump-layer connections to effectively capture crack micro-features and retain high-resolution spatial information. This allows the model to accurately segment crack boundaries, ensuring precise detection. Extensive experiments demonstrate that the model possesses strong generalization ability and delivers excellent test results on complex wall crack image datasets. By significantly improving detection accuracy, the U-Net model enhances the utility of automated inspection systems. This advancement holds substantial application value in building maintenance and safety assessment, contributing to more reliable and efficient infrastructure management.