Recently we experimented with producing a set of 3D models of items from the collection using a process called photogrammetry. Rose Attu, a Digital Humanities MA Student from UCL, who’s currently on placement at the BPMA, talks us through the process.
To create a 3D model a series of overlapping photographs of a single object are taken from different positions, and then a piece of software is used to triangulate the coordinates of each image and align them, reconstructing the whole object in 3D. With each shot, the digital camera also records metadata including the focal distance between the lens and the object; this enables the software to recreate the position the camera was in when each photo was taken.
The objects we chose were the prosthetic Postman’s Hand and the Stamp Snake. An object covered in a repeating pattern, or one without any distinguishing features at all, will affect the software’s ability to rebuild the depth of an object. Both these objects have plenty of distinctive features, so the software could detect the details from photo to photo and build accurate models without distortions.
Our first object was the Postman’s Hand. Once it was in frame and in focus, and evenly lit to avoid shadows, we could begin the capture process. To capture our objects from multiple angles we used a turntable to rotate them through 360°. The turntable was also covered in a distinctive image, which gave the software more common points to identify. The turntable was edited out at a later stage, so that our final models were just of the hand and the stamp snake.
On average it took 30 photos to complete one rotation, after which the camera was raised for a second sequence capturing the object from a higher angle. For more consistent results we kept the camera settings the same until a rotation was complete, and used a tripod and remote shutter release to keep the camera static. We output the raw image data as a set of TIFFs and corrected the white balance, and then our images were ready to be transformed.
Based on the estimated camera positions and the details in the images, the software built a point cloud, which is essentially a 3D model made up of dots. It then added a more detailed polygonal mesh layer representing the object surface. The final step was to add the textures; because we photographed our objects in high resolution, even the tiniest details were visible on the final models.
We were surprised with how little effort and technology it took, and the results were fantastic! For our set-up we used:
- Lazy Susan turntable with a nonrepeating pattern on it
- Digital SLR camera and tripod
- Light box or tent
- Studio lights or some source to help get rid of shadows
- Photogrammetry software – we used Agisoft which was great and you get a 30 day free trial
Have a go yourself– we would love to see your models!
– Rose Attu