Extraction of Open Spaces and Identification of Suitable Rooftops for Urban Agriculture: Contribution of Geospatial Technologies for Sustainable Planning

Urbanization represents a major challenge, driving the conversion of agricultural land into high-value uses such as residential, industrial, and commercial developments, largely due to rapid population growth. To address these challenges and enhance urban sustainability, it is essential to incorporate spatial planning strategies that integrate urban and peri-urban agriculture. We consider urban and peri-urban agriculture to be of vital importance in ensuring food security and meeting citizens' needs. Thus, our proposal is based on integrating agriculture into open spaces in urban and peri urban areas, such as community gardens, rooftops and greenhouses. In this way, we will be able to exploit these spaces for the cultivation of agricultural commodities common in the region, such as cereals and market garden crops. This project aims to extract those vacant spaces and rooftops using various methods. For vacant space extraction, three classifiers were used: Support Vector Machine (SVM), Minimum Distance, and Random Forest. With a 75% precision, the SVM classifier had the highest accuracy. For rooftop extraction, three methods were tested, including object-based classification using SVM, the pre-trained and optimized deep learning model Footprint Building Extraction-USA, and Mapflow's building model. According to our analysis, the last two approaches achieved the highest accuracy with F-Factor values above 77%.