Photogrammetry Remote Sensing and Spatial Information Sciences Research Illuminates Architectural Documentation

Strategic Insights -

High-resolution architectural documentation of historical buildings requires an in-depth understanding of their structure, materials, and historical layers, according to researchers from Polytechnic University Milan. A recent study implements a stone-by-stone segmentation approach using SAM2 to document the gothic complex of Milan Cathedral, leveraging object detectors for semantic interpretation. The research integrates 2D stone block segmentation with photogrammetric 3D reconstruction, enabling accurate projection of semantic labels and geometric data from images to 3D point cloud. This allows for a detailed 3D segmentation of all the components of the structure.

Key Takeaways:

  • Researchers from Polytechnic University Milan have developed a novel approach to high-resolution architectural documentation, involving stone-by-stone segmentation using SAM2.
  • The approach leverages object detectors for semantic interpretation and integrates 2D stone block segmentation with photogrammetric 3D reconstruction.
  • The framework enables accurate projection of semantic labels and geometric data from images to 3D point cloud, allowing for detailed 3D segmentation of all components of the structure.
  • The study focuses on the gothic complex of Milan Cathedral as the monument object of study.
  • The research concludes that recent advances in AI, particularly foundational models and zero-shot models, offer potential solutions to speed up the documentation process.

Statistics:

  • The study uses the SAM2 (Segment Anything Model) based stone-by-stone segmentation approach.
  • The research integrates 2D stone block segmentation with photogrammetric 3D reconstruction, enabling accurate projection of semantic labels and geometric data from images to 3D point cloud.
  • The detailed 3D segmentation approach allows for a precise understanding of the building's structure, materials, and historical layers.
  • The study provides a comprehensive framework for high-resolution architectural documentation of historical buildings.

Sources:

  • Milan Cathedral Digitized: A Stone-by-Stone Segmentation Approach Using SAM2. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2025, XLVIII-M-9-2025():1683-1690. (http://www.isprs.org/publications/archives.aspx)
  • doi: 10.5194/isprs-archives-XLVIII-M-9-2025-1683-2025
  • Copernicus Publications
  • Science Letter, October 24, 2025; p 1852