From Image to Map: Advancing GIS Decision Support Systems through Automated Building Detection and Vectorization Using the EURG DETECTIFY Interface

Deep learning-based image analysis technologies have significantly enhanced decision support systems across various fields, including Geographic Information Systems (GIS) and remote sensing. These technologies are essential in extracting features from remotely sensed images and are crucial for land-use planning, transportation infrastructure development, environmental monitoring, etc. Vector shapefiles of buildings can be foundational for various aspects of urban planning, such as land use planning and zoning decisions, infrastructure development, emergency response and management, environmental impact assessment, and many more. Also, detailed shapefiles with accurate geospatial information enhance urban development initiatives’ efficiency, precision, and public safety. Traditional methods of manually digitizing buildings in an image are time-consuming, costly, and less efficient. Despite showcasing promising outcomes, there are several challenges regarding postprocessing the output images of deep learning-based models, such as preserving the spatial information embedded in the input images. To deal with this issue, this study introduces a Graphical User Interface (GUI) called DETECTIFY, based on YOLOv8 deep learning model. This GUI streamlines the process of extracting building rooftops from images with and without spatial information. This also automatically georeferences the output image by transferring the spatial information from the input image and, lastly, generates precise vector shapefiles of buildings that can be used for visualization and analysis in any GIS software, enabling spatial information generation through simple, intuitive button interactions.