Urban building footprint and height extraction for enhanced urban analytics

Rapid urbanization has resulted in complex and heterogeneous cityscapes, making accurate information on building footprints and heights essential for urban planning, disaster management, and infrastructure development. Traditional methods, such as cadastral surveys and field mapping, are often costly, time-consuming, and lack scalability. Leveraging advances in deep learning and remote sensing, this study presents a fully automated pipeline for building footprint extraction and building height estimation from aerial imagery. The methodology integrates the ReFineNet model for precise building footprint segmentation and evaluates two complementary height estimation approaches: the MiDaS model for relative monocular depth-based heights and the nDSM method for absolute metric heights derived from DSM–DTM elevation differentials. The study compares these approaches in a medium-density coastal urban areas. Results demonstrate that both methods can effectively capture urban verticality, providing insights for urban density mapping, land-use analysis, and decision-making. The approach offers a scalable, cost-effective alternative to LiDAR or cadastral datasets for urban analytics and planning in data-scarce regions.