Automating the inventory of the navigable space for pedestrians on historical sites: Towards accurate path planning

Pedestrian mobility networks have a primary role on historical urban areas, hence knowledge of navigable space for pedestrians becomes crucial. Collection and inventory of those areas can be conducted by exploiting several techniques, including the analysis of point clouds. Existing point cloud processing techniques are typically developed for modern urban areas, which have standard layouts, and may fail when dealing with historical sites. We present a complete and automated novel method to tackle the analysis of pedestrian mobility in historic urban areas. Starting from a mobile lasers canning point cloud, the method exploits artificial intelligence to identify sidewalks and to characterize them in terms of paving material and geometric attributes. Output data are vectorized and stored in a very accurate high-definition shapefile representing the sidewalk network. It is used to automatically generate pedestrian routes. The method is tested in Sabbioneta, an Italian historic city. Paving material segmentation showed accuracy of 99.08%; urban element segmentation showed an accuracy of 88.2%; automatic data vectorization required only 1.3% of manual refinement on the generated data. Future advancements of this research will focus on testing the method on similar historical cities, using different survey techniques, and exploiting other possible uses of the generated shapefile.