Quantifying city dynamics: exploring the urban features representation of Milan’s streets

This study focuses on measuring, analyzing and representing the spatial distribution and correlations between urban visual features in Milan, Italy, within the context of the Multilayered Urban Sustain?ability Action (MUSA) project. Using Geographic Information System, Google Street View, and deep learning technologies, the research systematically analyzes streetlevel panorama pictures generated at sparse points in the city. Images are segmented according to urban categories ADE20K, focusing on greenery, ground, buildings, and sky. Through spatial continuity analysis, variograms reveal an aniso?tropic pattern, indicating significant visual continuity on specific orientations. The study discusses rela?tionships among urban features, such as the inverse proportionality between greenery and buildings/sky. Autocorrelation analysis confirm localized areas with similar feature values, while point-neighbor mapping identifies significant negative spatial correlations between greenery, buildings, and sky. The variograms illustrate maximum continuity ranges influenced by historical expansion processes, and shared continuity limit among all four categories. The uneven distribution of urban characteristics is evident in the heatmaps. The presented methodology can be adapted for similar analyses in diverse urban contexts, providing a valuable tool for urban researchers and planners.