Il Riuso delle linee ferroviarie per la mobilità dolce. Nodi critici e opportunità

Machine learning analysis of slow and short paths within dismissed railways tracks The spirit of this work is the overcoming of the dichotomy best path equals the shortest path, through the selection of alternative tracks that cross the territory with slow mobility, embracing the significant elements. The research in this chapter uses Image Segmentation tools to measure the spatial quality of slow mobility paths. This work is being developed within the research groups of DAStU - Politecnico di Milano: E-scapes Observatory and Mapfrag - Department of Excellence "Territorial Fragility". The methodology of this study aims to evaluate the quality of a path through tools and databases as widespread as possible, looking for solutions to a series of problems encountered. The case study in this chapter deepens the research of a process of optimization of the connec tion system of the Costa Verde of Trabucchi in Abruzzo using Image Segmentation technology. This is a subset of Machine Learning technologies used specifically for the process of partitioning an image in several polygons, typically used to locate objects and boundaries in images. Several tools for the analysis of paths through machine learning have been tested. Due to the selection made, Mapillary was chosen because it was more flexible for the needs of the research in progress. This engine was fundamental because, in addition to having the integration of an efficient Image Segmentation system, it allows the census of the most suitable areas for the development of slow routes, such as disused segments of railway networks. These areas are not included in Google's database, which is very efficient as far as roads are concerned, much less as far as ad hoc created cycle and pedestrian routes or greenways are concerned. Finally, a methodological experimentation of the different ways in which Mapillary acquires new images has been carried out, highlighting the difficulties encountered in the various surveying campaigns.