Automated façade image classification model for urban analysis in Murcia, Spain

Abstract

[EN] The expansion of cities in recent decades represents a significant social, demographic, and architectural phenomenon, underscoring the critical role of urban planning in achieving a balanced and sustainable approach to territorial development. Machine learning, a branch of artificial intelligence (AI), employs algorithms to extract insights from data, uncover patterns, and predict future outcomes. In the urban domain, while satellite imagery has long been the preferred choice for territorial analysis, facade images have recently emerged as a valuable resource. These images offer rich visual information, making them indispensable for studies ranging from identifying structural vulnerabilities to assessing urban vitality. This study combines Geographic Information Systems (GIS) and Deep Learning techniques with verified data from the General Directorate of Cadastre in Spain to analyze the urban environment using facade images from the city of Murcia, one of Spain’s most dynamic metropolitan areas. This study presents the development of an automated facade image classification model, implemented in Python using the pre-trained EfficentNetB0 architecture. This model, trained with cross-validation techniques, performs a binary image classification as part of a clustering analysis applied to the variables under investigation. The results were integrated into ArcGIS PRO, leveraging the cadastral references of the properties as key attributes for detailed spatial analysis. This approach enabled the identification of two significant areas linked to Murcia’s metropolitan evolution. In conclusion, the model successfully achieves the research objectives and serves as a foundation for more in-depth urban studies focused on the city of Murcia.