In the heart of Milan, a monumental task is underway, not with chisels and hammers, but with algorithms and data. Researchers led by K. Zhang from the 3D Survey Group at Politecnico di Milano are pioneering a new approach to architectural documentation, one that could revolutionize how we preserve and understand historic structures. Their work, published in ‘The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences’ (translated as ‘International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences’), is a testament to the power of artificial intelligence in heritage conservation.
The Milan Cathedral, a gothic masterpiece, serves as the canvas for this innovative research. The team is employing a cutting-edge AI model called SAM2 (Segment Anything Model) to achieve an unprecedented level of detail in architectural documentation. “We’re not just capturing the geometry of the building,” explains Zhang. “We’re delving deep into its structure, materials, and historical layers, interpreting the hidden construction logic, and identifying even the smallest components, like individual stones or bricks.”
This stone-by-stone segmentation approach integrates 2D stone block segmentation with photogrammetric 3D reconstruction. The result is an accurate projection of semantic labels and geometric data from images to a 3D point cloud. In simpler terms, it’s like giving the cathedral a digital twin, a detailed 3D model that includes not just the shape and size of every stone, but also its type, arrangement, and state of preservation.
The implications of this research extend far beyond the cultural sector. In the energy sector, for instance, similar techniques could be used to monitor and maintain critical infrastructure, such as power plants or wind turbines. Imagine being able to predict when a component might fail, just by analyzing its digital twin. This could lead to significant cost savings and improved safety.
Moreover, this research could pave the way for more efficient and effective conservation strategies. “By understanding the precise condition of each component, we can prioritize interventions and allocate resources more effectively,” says Zhang. This could be a game-changer for heritage conservation, making it more data-driven and less reliant on subjective assessments.
The team’s work is a remarkable example of how AI can be harnessed to solve real-world problems. As Zhang puts it, “We’re not just pushing the boundaries of technology. We’re using it to preserve our shared cultural heritage for future generations.” And in doing so, they’re opening up new possibilities for other sectors, including energy, to leverage these advancements for their own benefit.