PHOTOGRAMMETRIC SURVEY

David Lo Buglio (Université Libre de Bruxelles)


For the documentation of architectural elements, photomodelling (surveying, 3D modelling and image-based representation) has reached a point where the cost, accessibility, quality and diversity of the results are able to meet many of the needs and constraints of architectural surveying (technical and disciplinary). As a reminder, photomodelling is a 3D restitution technique based on photogrammetric image processing.

  • In a first step, each photographic image representing the object is oriented in relation to the others within a 3D reference space.
  • In a second step, the geometries of the object are modelled directly on the oriented photographs, which are then projected onto the geometries.

The projective relationship between the images is determined by the principles of epipolar geometry. The homologous (relevant) points of the object are identified on each pair of images, which makes it possible to cross-reference their positions by spatial triangulation. Once calibrated and oriented in space, the measurements and coordinates are extracted from the photographs. The projective correspondence established between the images and the geometry gives the opportunity to directly apply the texture of the object to the polygons.(1)

Recent advances in image analysis and computer vision make it possible to automatically process the phases corresponding to the extraction of homologous points, calibration, image orientation, the generation of dense point clouds and the reconstruction of polygonal meshes.(2) These developments of multi-stereo correlation make it possible for a real upheaval in architectural surveying.

The potential of approaches based on image-based 3D reconstruction (see "range-based metDAR") is becoming increasingly evident, particularly thanks to the latest developments in dense image correlation (3) and the availability of online or open-source tools (such as Photosynth, 123DCatch, Metashape, Apero, MicMac, etc.). These developments, based on photogrammetry and computer vision methods, have shown very promising results that make automatic photomodelling a low-cost, robust and practical alternative to 3D laser scanning.

The introduction of new spatial data acquisition and processing protocols is bringing new research perspectives. The possibility of envisaging the collection of massive data containing surveys, geometric analyses and semantic characterizations of heritage artefacts could also be used for new approaches to classification and comparison.

columnscolumnsPhotogrammetric survey. Multi-stereo correlation of convergent axis shots of a column base, Mauro Vincitore, in collaboration with the UMR MAP 3495, France.columnsColoured point cloud from the lasergammetric survey of the Romanesque columns of the cloister of Saint-Michel de Cuxa, France. In collaboration with the UMR MAP 3495.

(1) David Lo Buglio, Vanessa Lardinois, and Livio De Luca, ‘What Do 31 Columns Tell about a “theoretical” 32nd?’, ACM Journal on Computing and Cultural Heritage 8, no. 1 (2015), What Do 31 Columns Tell about a "theoretical" 32nd?.

(2) Marc Pierrot-Deseilligny, Livio De Luca, and Fabio Remondino, ‘Automated Image-Based Procedures for Accurate Artifacts 3D Modeling and Orthoimage’, in XXIIIth International CIPA Symposium, 2011.

(3) Heiko Hirschmuller, ‘Stereo Processing by Semiglobal Matching and Mutual Information’, IEEE Transactions on Pattern Analysis and Machine Intelligence 30, no. 2 (February 2008): 328–41, Stereo Processing by Semiglobal Matching and Mutual Information.

Vu Hoang Hiep et al., ‘Towards High-Resolution Large-Scale Multi-View Stereo’, in Proceedings CVPR 2009 (2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, USA, 2009), 1430–37, Towards High-Resolution Large-Scale MultiView Stereo.

Yasutaka Furukawa and Jean Ponce, ‘Accurate, Dense, and Robust Multiview Stereopsis’, IEEE Transactions on Pattern Analysis and Machine Intelligence 32, no. 8 (August 2010): 1362–76, Accurate, Dense, and Robust Multiview Stereopsis.

Anne-Laure Chauve, Patrick Labatut, and Jean-Philippe Pons, ‘Robust Piecewise-Planar 3D Reconstruction and Completion from Large-Scale Unstructured Point Data’, in Proceedings CVPR 2010 (2010 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), San Francisco, USA, 2010), 1261–68, Robust Piecewise-Planar 3D Reconstruction and Completion from Large-Scale Unstructed Point Data.