Fine-grained pixel-level crack mapping in earthen heritage structures.

The reliable smart tools that detect surface cracks in an efficient and accurate manner are needed for the heritage structure preservation. In this paper, an entirely automated framework is proposed in which high-resolution images have to be processed and relevant structural information and cracks are extracted in real-time. Image preprocessing, Skeleton-based analysis, and a lightweight deep learning model efficiently detect controlled and real environments in such a way that he workflow is integrated. An exclusive application was developed, through which users can also upload images, if needed, or snap shot images directly through a live camera so that the whole system can Then be applied in reality for practical site inspections. The model showed very high accuracy and very consistent performance under real-time on-field testing. Detection is not all that the system can do; scantling would also be offered on activities like crack width, average width, orientation, and severity. Thus, the proposed approach is just a proper and scalable answer to structural assessment while enabling a more rationalized action for maintenance decision-making in heritage management.