Trialling an Accessible Non-Contact Photogrammetric Monitoring Technique to Detect 3D Change on Wall Paintings

ABSTRACT

The visual comparison of 2D images, or ‘photographic monitoring’, is commonly used in cultural heritage conservation to assess rates of deterioration and the impact of interventions. While it is an accessible technique, three-dimensional changes can be difficult to discern and identifying change can be subjective. Structure from Motion (SfM) photogrammetry, which generates 3D point clouds from digital photographs, has the potential to make monitoring more objective and more informative by recording three dimensions of change rather than two. A methodology was tested on the Romanesque wall paintings of St Botolph’s church, Hardham, UK to develop a SfM photogrammetric condition monitoring system that is affordable, portable, non-contact, and repeatable. Commercially available camera equipment was used for capture, and readily available or free software was used to generate and detect change between two 3D point clouds. Time-SIFT, an in-software workflow of point cloud alignment, was used as a reliable non-contact method to replace adhered or embedded registration points on the wall to align the clouds from different capture epochs. The technique was able to identify change in the wall paintings, with the smallest verified changes at a scale of 0.5 mm. The results are discussed in relation to their advantages over 2D imaging, the avenues for further research, and the limitations of applying the technique on site. This research demonstrates the potential of photogrammetry as an accessible in situ condition monitoring tool using a workflow and equipment tailored to working on sites with wall paintings.

KEYWORDS:

• Photogrammetry
• cultural heritage
• condition monitoring
• wall paintings
• change detection
• imaging
• accessibility

Acknowledgments

The authors would like to thank Professor Jim Chandler of Loughborough University for his assistance to this research, Lori Wong for her invaluable feedback and encouragement in publishing this work, Luigi Barazzetti for his patience, expertise, and time, and Valerio Sabbatini for his support in its dissemination.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Software: Image Comparer: https://www.bolidesoft.com/imagecomparer/photo-comparison.html (accessed 11/01/2020); National Gallery developed software VIP-ip: x https://github.com/libvips/libvips (accessed 11/01/2020); also National Gallery developed software IIIPImage: https://iipimage.sourceforge.io/ (accessed 11/01/2020). For a literature review on 2D change detection applied to cultural heritage, see Abate 2019b.

2 Agisoft Metashape Professional has two advantages. Targets are automatically detected producing high precision results, and the epochs can be scaled together without exporting to another software. The lower cost Agisoft Metashape Standard could be used if desired with additional software to scale the point clouds in other software. Manual: https://www.agisoft.com/pdf/metashape-pro_1_5_en.pdf. Other SfM MVS software options are available (e.g. Reality Capture, Photomodeler Pro 5, 3DF Zephyr), but Agisoft Metashape is a relatively affordable, user-friendly option, with the necessary functionality.

3 https://www.cloudcompare.org/doc/wiki/index.php?title=Main_Page

4 A 50 mm prime lens produces a field of view that is close to the angle of the human eye when paired with a full frame sensor, and is recommended by the software (Verhoeven 2016; Agisoft 2019). Due to the strength of Agisoft Metashape’s lens distortion calculations, a different focal length can be used providing the equipment remains the same throughout the monitoring programme. Aperture f/8 allowed for a maximum depth of field range of 2 cm at its minimum focal distance of 45 cm and is the smallest aperture that can be used before excessive diffraction is induced (see Nikkor Manual at https://downloadcenter.nikonimglib.com/en/products/298/AF_Nikkor_50mm_f_14D.html). While a smaller aperture of f/11 would increase the depth of field, it would result in image-distorting diffraction, less light on the sensor and therefore a longer shutter speed, and increased risk of shake or blur. Similarly, a larger aperture of f/5.6 would allow for more light to enter the camera, a faster shutter speed, and less diffraction, but would further limit the depth of field of the image, meaning part of the image may be out of focus. There must always be a balance between depth of field, distance from the subject, and image sharpness.

5 This was the minimum focus distance for this chosen lens; however it is possible to purchase 50mm lenses that have an even smaller minimum focus distance.

6 Using Agisoft PhotoScan Pro v1.2.6 (now Agisoft Metashape)

7 PLY or Polygon File Format is a file format designed to store 3D data from 3D scanners, and is able to store the basic xyz coordinate information, along with scale and detailed colour and texture information. Turk, G. http://paulbourke.net/dataformats/ply/ (accessed 2021/04/05).

8 Octrees are trees of data points, with each ‘node’ (or level of data) having eight subdivisions. Each node represents a volume entrained by a cube. The cube is then subdivided into eight more cubes, with each octree level breaking the cubes into small subdivisions (Figure 6). Once the subdivision ends, each node should be theoretically uniformly occupied (i.e., with a minimal number of points) (Elseberg et al. 2012). Using octrees, the points are assigned coordinates which make it easier to find their nearest neighbour in the reference cloud.

9 These are large scale changes which can visually overwhelm the smaller scale changes of interest. The scale of change can be manipulated to highlight smaller scale changes, or the area can be cropped.