This is the third post in a series about this decorated box, which I started back in March, and then continued by trialling some photogrammetry through the mobile app version of RealityScan. Here I’m attempting the same box but using the desktop application, DSLR camera and basic home studio setup with makeshift lighting.

The purpose of this project is to practice photogrammetry techniques and equipment before I venture into the field to document real cultural heritage objects. The aim of this trial is to obtain precise accuracy in fine detail and a true representation of the object as a 3D mesh. As previously, the result is shared here with SketchFab as a running constant, but acknowledging its limitations as an online viewer.

This result is achieved from 340 images, around double the total of the mobile app, but taken in the same way using a turntable with 10 degree increments. The camera is a Canon 5D mark III, recording 22.3 megapixel images at double the resolution of my camera phone used in the previous attempt. The lens is a standard 24-70mm zoom lens without any modification, and taking a range of images from various distances and heights as close as the focus would allow. To light the object, I have used a low-budget combination of household lamps with translucent shower caps as diffusers to produce even lighting. A homemade 250mm scale bar and Colorchecker target were also included in the first shots for calibration.

Since I last used the software a few years ago, RealityScan has now introduced a new payment model that makes it free to use for small businesses, educational institutions and students. It also has a rich archive of tutorials and inbuilt help pages to lead through the process step by step, from registration of the images and processing the mesh, to texturing the model.

First the images were processed from RAW to JPG through Lightroom, also balancing the exposure levels for minimum shadow/highlights and verifying the colour before loading into RealityScan. Using the normal detail setting, my first model came out with 14.5 million polygons, and all 340 images registered. Once textured, the images combined to produce a satisfactory albedo texture – an even base colour texture without strong lighting contrasts to which simulated lighting can be applied.

This process is more labour intensive than the mobile app, but gives more control over the result. At maximum resolution, for the most part this model appears to have captured a high level of detail at normal viewing distance. The software allows for automatic export to SketchFab, but I discovered it was best to decimate or simplify the mesh first, where there was greater control in the detail that had to be lost. Once under 1 million polygons, the model uploaded with a reasonable level of perceivable quality.

Overall this was a successful capture, but not without some anomalies and slight noise in some areas. Some images came out blurred, where better lighting and control over close range focus could have helped. I am also interested in capturing tool marks and close up detail, recording the craft and artefacts of production as much as the object itself (for more see my articles looking at the work of Factum Foundation). That might require venturing into macro photography, but within the limitations of this particular home studio setup the result was quite satisfactory, for now.