Artificial intelligence applied to cadastral management: development and implementation of a web-based analysis system

Abstract

Urban cadastral management is a central challenge for local governments, as its accuracy and up-to-dateness are crucial for tax equity, territorial planning, and public policy decision-making. In many Latin American countries, cadastral records exhibit significant backlogs or discrepancies between what is recorded and what has been built, affecting tax collection, institutional transparency, and efficient land use (Erba and Piumetto, 2021). In this context, technologies based on artificial intelligence and computer vision, particularly semantic segmentation applied to remote sensing images, have demonstrated the capacity to automate the detection of built-up areas, mitigate errors in physical inventories, and facilitate frequent updates (Pluto-Kossakowska et al., 2025). Recent models such as YOLOv11 have been successfully employed to segment objects and buildings in urban and construction environments, achieving good accuracy rates (mAP) in applied studies (He et al., 2024). This paper presents the development and implementation of an interactive web system that combines semantic segmentation techniques with geoprocessing and visualization modules. The browser-based application allows users to upload aerial images, compare AI results with official records, and generate georeferenced reports of discrepancies between built and declared properties, for practical use in cadastral management, with the potential for nationwide scalability.