AIC's 53rd Annual Meeting

The ability to generate 3D models of heritage objects opens up many exciting possibilities for conservation. However, the options for 3D digitizing objects are either prohibitively expensive for many cultural institutions (structured light scanning and laser scanning), or too slow and skill-dependent to be of use in scanning large collections (photogrammetry). In an effort to address the limitations of photogrammetry as a commodity 3D digitization technique, a game programmer and a conservator began collaborating to reduce the total time for producing a model to ~10 minutes.

This paper presents their interim results on doing photogrammetry quickly, accurately, and reproducibly using an automated turntable and multi-camera arm along with Agisoft Metashape Pro and a processing pipeline written in the Python programming language. 

The pipeline script waits for photographs from a multi-camera, processing them and building masks while photos continue being taken. A palette of computer-readable markers placed beneath the object on the turntable allows for scaling and the orientation of the object in space such that each model is oriented and scaled consistently in relation to the others. When photography has completed, the images are automatically built into a model using Agisoft Metashape Pro. The finished model is exported, and an archival render of the 3D model is taken using Blender.  

After initial setup and configuration, the pipeline requires little further user input and can build a model in as little as ten minutes and thirty seconds from starting photography to the export of the final model.  The pipeline script is extensible, configurable, and is usable with or without an automated turntable. This  method promises to make photogrammetry not only faster, but more efficient, consistent, and accessible to a wider number of institutions.