Structure from Motion

There are always so many things to do and learn about! Its fun having Jordan Mertes here in Innsbruck for a while, as he is expert in using Structure from Motion tools to develop surface terrain models. I’ve dabbled in this a fair bit but usually feel a bit out of my depth and am not always sure if I have generated models of the best possible quality.

The principle is that you take a bunch of photos of an object or surface of interest, from a variety of angles and distances and specialized software can be used to identify a large number of common features in the images, perform a suite of positional calculations on these and determine where the photos was taken from and also the 3D form of the surface being photographed. Its pretty cool stuff. Anyone can do it! Like many colleagues of mine, I use Agisoft Photoscan software, which you must pay for, but there are other freeware versions using the same principles such as Visual Structure from Motion. All you need is a camera and a computer and a natty idea …

Ben Partan and I took a bunch of images and videos in 2013/2014 to see how well we could use this  kind of photogrammetry and software to create surface models of snow penitentes in Chile to compare with the models we generated using the Microsoft Kinect as a close range 3D camera (see Nicholson et al., 2016).

Penitentes are a hard challenge for terrestrial photography as the contrast is poor and the surface is complex and shadowed. It was a very good learning ground for us – we could see how well different configurations of photo vantage points compared to extracting frames from videos, and we tried both cheap and high quality cameras.

Below is an example model of snow penitentes in a river bed, where we tested all our equipment and set up. The image is a screenshot from Agisoft Photoscan software. You can see from the camera positions indicated in blue that we were trying two approaches: (1) taking an array of photos from a single location and moving on to do the same at another, and (2) taking a single photo, and then moving along a meter to take the next one.

SfM_penitentesYou can read much more about Structure from Motion (SfM) applications in earth science in this new book: Structure from Motion in the Geosciences by Jon Carrivick, Mark Smith and Duncan Quincey, and in Westoby and others (2012). A very timely publication indeed as these techniques are booming in earth science and perhaps especially in glaciology at present. You can read more about the penitentes we measured in our publication in the Cryosphere.

Westoby, M. J., Brasington, J., Glasser, N. F., Hambrey, M. J., & Reynolds, J. M. (2012). “Structure from-Motion” photogrammetry: A low-cost, effective tool for geoscience applications. Geomorphology, 179, 300–314. doi:10.1016/j.geomorph.2012.08.021

Nicholson, L. I., Petlicki, M., Partan, B., and MacDonell, S. (2016) 3D surface properties of glacier penitentes over an ablation season, measured using a Microsoft Xbox Kinect, The Cryosphere, 10, 1897-1913, doi:10.5194/tc-10-1897-2016, 2016.

About lindsey

Environmental scientist. I am glaciologist specialising in glacier-climate interactions to better understand the climate system. The point of this is to understand how glaciated envionments might change in the future - how the glaciers will respond and what the impact on associated water resources and hazard potential will be.
This entry was posted in education, my research. Bookmark the permalink.

Comments are closed.