METHODS FOR AUTOMATING THE CONSTRUCTION OF GEOSPATIAL OBJECTS' 3D MODELS

Modern methods for creating digital models of geospatial objects are based on the integration of conventional photogrammetric techniques and unmanned aerial vehicles (UAVs) with advanced artificial intelligence algorithms for processing multiple overlapping images. This integration significantly improves the quality and efficiency of constructing 3D models of geospatial objects. Structure-from-Motion (SfM) and Multi-View Stereo (MVS) algorithms are effectively complemented by deep neural networks (Deep Learning), enhancing the 3D modelling process, including the reconstruction of challenging surfaces. It has been established that these methods are most effective when sufficiently large training datasets and high-performance computing resources are available. This highlights the need for the further development of cloud-based services and their capacity to process large volumes of geospatial imagery in real time. Potential directions for further research aimed at facilitating the large-scale implementation of automated 3D modelling methods are also outlined.