A Deep Learning Approach for the Recognition of Urban Ground Pavements in Historical Sites

A deep learning approach achieved 94% accuracy in classifying five types of urban ground pavements—sampietrini, bricks, cobblestone, stone, and asphalt—in the historical site of Sabbioneta, Italy. The method leveraged a Mobile Mapping System (MMS) to capture point cloud data, which was then rasterized and analyzed using a U-Net model coupled with ResNet 18. This high-precision classification offers a promising tool for automating pavement mapping in historical urban areas, supporting sustainable city management and smart urban planning.

Treccani, D., Balado, J., Fernández, A., Adami, A., & Díaz-Vilariño, L. (2022). A DEEP LEARNING APPROACH FOR THE RECOGNITION OF URBAN GROUND PAVEMENTS IN HISTORICAL SITES. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLIII-B4-2022, 321–326. https://doi.org/10.5194/isprs-archives-xliii-b4-2022-321-2022