Abstract
Abstract. 3D detailed models derived from digital survey techniques have increasingly developed and focused in many field of application. The high detailed content and accuracy of such models make them so attractive and usable for large sets of purposes in Cultural Heritage. The present paper focuses on one of the main techniques used nowadays for Cultural Heritage survey and documentation: the image matching approach or Structure from Motion (SfM) technique. According to the low cost nature and the rich content of derivable information, these techniques are extremely strategic in poor available resources sectors such as Cultural Heritage documentation. After an overview of the employed algorithms and used approaches of SfM computer vision based techniques, the paper is focused in a critical analysis of the strategy used by two common employed software: the commercial suite Agisoft Photoscan and the open source tool MicMac realized by IGN France. The experimental section is focused on the description of applied tests (from RPAS data to terrestrial acquisitions), purposed to compare different solutions in various featured study cases. Finally, the accuracy assessment of the achieved products is compared and analyzed according to the strategy employed by the studied software.
Highlights
Dense image matching methods enable the extraction of 3D point clouds and the generation of 3D models through a processing of a set of unoriented images acquired from multiple views
The present paper focuses on one of the main techniques used nowadays for Cultural Heritage survey and documentation: the image matching approach or Structure from Motion (SfM) technique
After an overview of the employed algorithms and used approaches of SfM computer vision based techniques, the paper is focused in a critical analysis of the strategy used by two common employed software: the commercial suite Agisoft Photoscan and the open source tool MicMac realized by IGN France
Summary
Dense image matching methods enable the extraction of 3D point clouds and the generation of 3D models through a processing of a set of unoriented images acquired from multiple views. While traditional photogrammetry derives calibration parameters of the camera and the camera poses mainly from well-distributed GCPs and tie points, a Structure from Motion (SfM) approach computes simultaneously both this relative projection geometry and a set of sparse 3D points To do this, it extracts corresponding image features from a series of overlapping photographs captured by a camera moving around the scene (Verhoeven et al, 2012). The second category represents surface reconstruction methods, where dense image matching algorithms exploit the previously derived orientation of the images to derive complete surface These techniques allow the generation of 3D information even if the images are acquired by non-expert people in the field of Photogrammetry and 3D reconstruction (Pierrot-Deseilligny et al, 2011). From the gradient orientations of sample points within a region around the keypoint in order to get an orientation assignment (Lowe, 2004 ; Ke and Sukthankar, 2004)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
