Skip to Main Content
This paper presents an automated methodology for the accurate reconstruction of surfaces and 3D objects. This method was developed for close-range photogrammetric applications, with a particular attention to terrestrial free-form objects that can be modeled with point clouds extracted from high resolution images. The image orientation phase is carried out using an automatic procedure based on a rigorous bundle adjustment. Then, an approximate model is obtained using a patch-based matching strategy. Then, this initial model is refined using a Multi-Image Least Squares Matching (MGCM) implementation, increasing in this way the precision of the final model and giving the possibility to evaluate its consistency using a rigorous variance-covariance propagation. The multi-image matching result is a dense point cloud describing in an accurate way the object surface. The derived point cloud is used to obtain a mesh which is finally textured obtaining a photo-realistic result.