Digital Archive from the Analysing Britain's Most Elusive Roman Sculptures Project
Ian Haynes, Lindsay Allason-Jones, Alex Turner, 2023. https://doi.org/10.5284/1090382. How to cite using this DOI
Data copyright © Elusive Sculptures Project unless otherwise stated
This work is licensed under a Creative Commons Attribution 4.0 International License.
Primary contact
Ian
Haynes
Professor of Archaeology
Newcastle University
School of History, Classics and Archaeology
Newcastle upon Tyne
NE1 7RU
UK
Resource identifiers
- ADS Collection: 4457
- DOI:https://doi.org/10.5284/1090382
- How to cite using this DOI
Overview
Elusive Sculptures – reflection on methodologies – Alex Turner
The capture of data for Elusive Sculptures employed, alongside conventional 2D photography, two 3D methods of data capture. Structured Light scanning was undertaken using three different instruments from the company Artec3D, alongside Artec Studio software and Structure from Motion (SfM) was employed using a conventional digital SLR and Agisoft Metashape or Autodesk Recap Photo as the processing software. What follows is a brief discussion of the relative merits and demerits of each method.
Conventional digital photography
Conventional photography is a tried and trusted method of recording sculpture and has the advantage that it is easy to deploy and can, with additional lighting equipment, produce subtlety detailed images of each object. If the white balance of each image is carefully adjusted to the local lighting conditions, it is also possible to produce accurately rendered colours for each object. Even if the final publication is printed in monochrome, the colour data is still held as a valuable resource within the photographic archive. However, the weakness of digital photography lies in the difficulty of capturing objects in inaccessible or architecturally crowded places. Whilst it is perfectly possible to edit out a cluttered background in Photoshop, this is a time consuming and sometimes less than satisfactory process. Additionally, even with a fixed focal length lens, the view that is presented within a photograph is a perspective rather than an orthographic one and some distortion is inevitable. Whilst this may give the subject a more aesthetic appearance, it is difficult, even with a visible scale, to take accurate measurements from the printed output.
Structure from Motion
Structure from Motion (SfM), sometimes referred to as photogrammetry, is a method by which a series of 2D photographs is taken with a fixed focal length lens and then processed using software that employs an algorithm to identify common points between photographic pairs. This geometric information is then used to derive a series of points in 3D space that represent the object as a basic point cloud. This point cloud is further refined by the creation of a dense point cloud that is generated from the geometric interpolation of the basic cloud. A solid 3D model is generated as an irregular triangulated network based on the points (vertices) within the dense cloud. A final texture for this model is created by calculating the colour of each vertex and triangular face within the model and creating a correctly mapped surface texture.
There are several advantages of using this methodology but there are also a number of problems that need to be mitigated when using this approach. The major advantage over conventional photography is that the correctly scaled model is produced as a 3D object that can be interrogated to produce a much wider range of detailed measurements that can also include calculation related to volume and weight. It is also possible to derive cross-sections through any model created and given a suitable platform share these models as interactive manipulatable objects. As the model is derived from a conventional digital SLR, given the correct lighting conditions, the colour rendition of the surface texture should be accurate.
This methodology is, however, difficult to deploy when larger objects are placed in small, cluttered spaces or where access to parts of the object are not possible. Several of the objects that were captured, as part of Elusive Sculptures, are only modellable as single faces. Many of the fuller 3D models lacked data capture from the rear of the object, as in many cases this was simply inaccessible due to the proximity of walls or other objects. The use of SfM is also particularly difficult in architecturally crowded environments where similarly textured objects surround the subject being recorded. This can lead to the algorithm within the software misidentifying matches between photographic pairs and the production of models that morph into their immediate surroundings. Unlike object SfM, within a carefully controlled studio space, the real world is much more challenging and it is not always possible to remove background clutter or create the correct lighting conditions.
Structured Light scanning
Structured Light scanning was conducted on a large number of objects as part of Elusive Sculptures with a varying degree of success. Three scanners were used, the Artec Eva, Artec Spider and the Artec Leo. All these scanners employ video frames captured through a series of lenses that are arranged is such a fashion as to enable the software to calculate the 3D position of the surface of an object in real time and translate this into a triangulated network. Strobe lighting is built into the instruments to provide local illumination of the object.
Post-processing of the data is required with the Eva and the Space Spider, whereas the more recent Leo is able to do much of the processing in real time, dramatically reducing the time taken to produce the final model. The majority of the early captures were done with the Artec Eva that required a connection to a laptop at the point of scanning and was therefore physically less flexible than the models acquired with the self-contained Artec Leo. Both scanners have a maximum resolution of 0.2mm and are capable of capturing objects up to approximately 3m in size. The Artec Spider has a maximum resolution of 0.1mm and is only suitable for the capture of smaller objects. In practice, the Spider was only deployed on a few occasions and was the least successful in terms of quality data capture. The major drawback with these instruments is that they require close physically proximity, between 17cm and 170cm, to an object and that they must have a physical view of the surface they are recording. As with SfM, all round access to objects in a crowded museum collection is problematic. Where complete access to an object was impossible the emphasis was on the capture of the sculptural detail of an object and even though incomplete as a 3D model, it was still possible to produce orthographic images from the resultant output. The instruments use of multiple lenses aligned at specific angles allowed the easy extraction of objects from cluttered backgrounds during post-processing, a daunting task in both conventional photography and SfM data capture.
In common with SfM, the models produced from these instruments are metrically correct and measurements taking from them have millimetric accuracy. The biggest shortcoming of this method of data capture is the limited control of colour balance during scanning. It is noticeable that where SfM and Structured Light scanning were employed on the same object, the colour of the generated texture was much more accurate with SfM.
