This study introduces a new framework for combining calibrated mobile mapping system (MMS) data and low-cost unmanned aerial vehicle (UAV) images to generate seamless, high-fidelity 3D urban maps. This approach addresses the limitations of single-source mapping, such as occlusions in aerial top views and insufficient vertical detail in ground-level data, by utilizing the complementary strengths of the two technologies. The proposed approach combines cloth simulation filtering for ground point extraction from MMS data with deep-learning-based segmentation (U²-Net) for feature extraction from UAV images. Street-view MMS images are projected onto a top-down viewpoint using inverse perspective mapping to align diverse datasets, and precise cross-view alignment is achieved using the LightGlue technique. The spatial accuracy of the 3D model was improved by integrating the matched features as ground control points into a structure from the motion pipeline. Validation using data from the campus of Yonsei University and the nearby urban area of Yeonhui-dong yielded notable accuracy gains and a root mean square error of 0.131 m. Geospatial analysis, infrastructure monitoring, and urban planning can benefit from this flexible and scalable method, which enhances 3D urban mapping capabilities.

Road information high definition maps (HD map) contain information about the facilities around the roads and are often constructed through a mobile mapping system (MMS). Although constructing an HD map is essential for road maintenance and the application of autonomous driving in the future, it is problematic to acquire the data of objects other than the facilities in an unstructured form while operating the MMS. In this study, the researchers define this object data as clutter objects and present a method of automatic removal using characteristics of the MMS and image segmentation techniques. By applying the method to 10 KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute at Chicago) datasets, clutter objects were removed with an average overall accuracy of 91% with 0% (0.448%) error of commission for the complete point cloud map.