Deep Learning methods and geographic information system techniques for urban and territorial planning

In the urban planning context, the integration between Artificial Intelligence (AI) and Geographic Information System (GIS) assumes notable importance. Following these assumptions, this research aims at developing an integrated platform which, through Deep Leaning (DL) techniques, automatically classifies and segments the building heritage and the Urban Green Areas (UGAs) from satellite and aerial images. The obtained results are imported in the GIS system to identify the spatial relation between the segmented urban components. To assess the transferability of the developed platform, the extracted buildings and UGAs are imported into the GIS system and transformed into vectorial layers. The authors applied the method in the Municipality of Rende (Italy). The obtained results show a good ability of the model to identify and predict the existing buildings and UGAs in the investigated urban system. This integrated platform represents a useful Decision Support System (DSS) for the management and planning of different urban and territorial components, as well as a useful tool for updating urban data which, most of the time, are incomplete and inconsistent.