What is Photogrammetry?

Photogrammetry is a technological technique using a camera and modeling software to replicate something digitally in 3 dimensions. The technique itself can be used to create 3D digital models of just about anything, but for our use in Archaeology, it allows us to document and share artifacts and even whole excavation trenches. The technique itself is simple:

First, take photos in a sphere around the object. This can be executed using any kind of camera, however better resolution is obviously desired. It is also more desirable to use a camera on which you can change lower the ISO as low as possible and raise the F-stop to as high as possible. This makes for more light in each shot, which only helps the software analyze each photo.

Second, import the photos into photoshop to make image “masks” to be used on photoscan. These masks help the software identify what object or area you want to be modeled. Making a masks pretty much just means highlighting the area you want to be modeled, while deleting the rest of the picture. In our case, for the trenches, we just had to draw a square around the trenches, then tell photoshop to delete the rest of the picture. Save these edited images somewhere easily accessible for the last step.

Lastly, use your desired software to create a 3D model using your edited pictures. We used Agisoft Photoscan, but there are other programs that work as well. Once you upload all of the pictures into the software, you have to align all the photos. The software can automatically do this by identifying identical pixels in each picture and orienting the object based on these similar pixels. Once the photos are aligned, a dome of camera angles will appear, and you will have a sparse image cloud. (See below Figure 1):

(Figure 1: a picture of Trench 3 with cameras overlaid)

The sparse cloud will end up being a bunch of tiny dots that are roughly in the shape of the object. These dots are pretty much individual pixels that the software identified and placed in the form of the object. Some of these dots are going to end up floating in space, and there will be many other extra dots that you won’t need because they aren’t part of the object. After deleting all of these dots, you will be able to create a dense cloud of your image, which is basically a sparse cloud but with much more points. After touching up the dense cloud as well, you can then create a mesh, which creates polygons from all of the points and forms a solid object that should resemble your object (see below Figure 2).

(Figure 2: a picture of Trench 3’s mesh in Agisoft Photoscan)

It is interesting to note that the polygons created are in fact triangles. We learned from Austin Mason that these are used because it is the only polygon that can’t be warped (two sides of a rectangle can be skew). However, a triangle is unambiguously in a plane: three points define a plane. This was an interesting geometric insight into how photogrammetry works.

Note: When 3D modeling an object like a trench, where you only need to see the top, you can create these clouds very easily using the one “chunk” that makes up your sparse image cloud. When modeling an artifact such as the piece of concrete we modeled, we needed to create two chunks. One from about the middle of the object going to the top, and one from the middle of the object going to the bottom. All in all, we had to do everything twice before we are able to auto-merge the two chunks.

When you have a mesh of your object, the only thing left to do is to create a texture for the mesh, which is just an image that is stretched over the mesh. Once again, the software can do this for you using the data from the photos you gave it.

The only thing left to do is upload the model when it’s ready! Sketchfab is a website that can make publishing your models very easy. We used Sketchfab to embed our models on this website. Just click on the play button over the image and it the model will load; momentarily, you’ll be able to click and drag to rotate the 3D objects! Have a look:

This is a model of Trench 1. Note: We forgot the north arrow and scale, but the red end of the fire hydrant is pointed approximately West and the trench area dimensions are 1 meter X 1 meter.

This is a model of Trench 2, again its area dimensions are 1 meter X 1 meter.

This is a model of Trench 3, we do not have data on its dimensions.

This is a model of a cement artifact from Trench 2. It could possibly be a chunk of some foundation of the temporary buildings that made up Pine Hill Village.

Conclusion:

One of the main concerns of this project was how it would contribute to the overall Pine Hill Village Archaeology project. True enough, photogrammetry doesn’t reveal anything that isn’t visible to the naked eye viewing the real object, however, one other thing of value of the technique of photogrammetry is the quality of preservation. Although we had Community Archaeology Day, it was only one day and now it has passed. Additionally, to conclude the class, we backfilled the trenches. But with the models that we made, the excavation trenches (and one artifact) from this project will be preserved and sharable online with the general public permanently. There is something to be said for permanent preservation using zero physical space–especially given the storage crisis Archaeology as a discipline faces today.

One thing that we had trouble doing was keeping the scales in our models. The first few times, we forgot to include scales and north-arrows in the photos, but even after we added scale bars, upon merging the chunks, the scale bar was extirpated and thus not represented in our models. We did, however, have more luck exporting and integrating the models onto the webpage than previous years. We went through a third-party program, called Sketchfab, that has seamless embedding with WordPress. We highly recommend this option.

Finally, for future use of photogrammetry, we would recommend modeling objects that won’t last forever, like excavation trenches or decaying/fragile artifacts. Otherwise, the benefits of photogrammetry are reduced to storage optimization. One might also consider using photogrammetry strictly for important artifacts that’s shape is particularly interesting and difficult to describe in words. If the shape of the object is of particular interest, for instance, a bowl that’s shape refers to a specific time period, then photogrammetry is all the more helpful. One thing to avoid, unless a solution is considered, is photogrammetrizing glass. The problem is opacity and reflection. This can be resolved with masking spray or well-thought-out lighting.

Hopefully Carleton students will continue to use photogrammetry in the coming years. There’s a lot of potential for modeling interesting artifacts.