Web-Sourced Multi-Resolution Drone Imagery for Structure-from-Motion Modeling

This article presents the application of structure-from-motion (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{S f M}$</tex>) techniques to process multi-resolution drone imagery and drone-like imagery, with reconstructed results based on web-sourced multi-resolution images for SfM modeling. With the increasing accessibility of drones, highresolution aerial imagery plays a vital role in landscape modeling and visual storytelling. Challenges arise when handling images with inconsistent resolutions and viewpoints, often leading to fragmented models and visual distortions. This article examines the impact of extracting feature correspondences across images of varying resolutions and estimating camera poses to generate a sparse point cloud. The process continues with dense matching and surface reconstruction using Poisson surface reconstruction to generate triangular meshes. The complete SfM pipeline encompasses image capturing, feature extraction, feature matching, sparse bundle adjustment, dense reconstruction, and mesh generation. This framework improves the usability of diverse drone datasets and supports the advancement of web-based spatial visualization.