What’s in a photograph?

By Lucy Clarke (@DrLucyClarke)

Everyone is familiar with photography, with the rise of digital cameras and increasingly high resolution cameras available on mobile phones and tablets people are photographing everything from their holidays, pets, family, friends and even themselves. I love photography and enjoy capturing images of my travels, but I also use photographs in a very different way as part of my research: rather than just appreciating their aesthetic value I use photographs to recreate and measure features on the Earth’s surface and in the lab.

In remote areas where it is difficult to access a location or when looking into the past, photographs can often be the only option available to explore an area. Historic photographs are therefore extremely valuable, as they provide a record of what something looked like at various points in time and so can be used to look at temporal change. This is especially useful if an extreme event occurs in a location that has never been measured before, so you can look at the impact the event had. An example of this is shown below; the 2 photographs show the Poerua alluvial fan in New Zealand before and after a big event. In 1999 a large rock avalanche occurred in the headwaters of this system forming a dam in the gorge below, the water ponded up behind this for 2 days before it finally burst, creating a flood wave that engulfed the area downstream and caused the river to avulse (move to a new location) and deposit large areas of gravel on top of the agricultural land. Using photographs from before and after this event enables identification of the area of land that has been affected and the new position of the river channel to try and assess the damage.

Aerial photographs from 1984 (before) and 2005 (after) the 1999 flood event on the Poerua alluvial fan in New Zealand (Images courtesy of NZ Aerial Mapping Ltd and GeoSmart)

Although it is useful to look at these photographs and see the changes between when they were taken and what is there now, it doesn’t help us to actually measure anything or quantify the change. So in my research I use a technique called photogrammetry, which allows me to process photographs and extract quantitative data from them. In its simplest form, photography converts the 3D real world into a 2D image, and photogrammetry converts this 2D image back into a 3D representation - using information on the type of camera and lens used to take the image and the relationship between the camera and the ground at the time that the image was captured. This requires two overlapping images of the same place which are viewed at the same time in a single 3D image, in what is known as a stereo-image. Traditionally, this was done using a stereoscope (which uses mirrors and viewing lens to fuse the 2 images together when you look at them - like a magic eye picture) but in modern digital photogrammetry this is done using specialist software on a computer using a 3D screen and glasses – like when you watch a 3D film at the cinema. In the digital workflow the images are adjusted according to the camera parameters and georeferenced using the coordinates of known positions from the ground to create a true scale representation. A digital elevation model (a 3D map of the surface area) can then be extracted and used to measure features, this gives the same results as it would have done if you were standing on the ground measuring it.

Ways of viewing images in stereo (a) the traditional method using stereoscope and (b) my digital photogrammetric computer set up with 3D screen and glasses

Photogrammetry is most commonly used with aerial photography but it can be applied to any overlapping imagery if you have the correct information. For example, below is a photograph and associated digital elevation model I created from my alluvial fan experiments, outlined in my previous blog post: What drives change on alluvial fans? 

Photogrammetry software is expensive to purchase and processing the images can be complex and involves training, so traditionally photogrammetry has only been used by specialists. But recently there has been a development in something called Structure for Motion – this involves taking multiple photographs of objects from different angles and then uploading these into software that uses photogrammetric principles to automatically create a 3D model. This software is available on the web – e.g. Bundler (free to download), Photosynth (free to download) and AgriSoftPhotoscan  (the demo version is free, which allows you to create models but not save) - so you can upload your own photos and have a go at creating your own 3D model from them!

An example of the 3D model created by Structure for Motion (Source: Goesele et al, 2007)

I first used photogrammetry many years ago for my Masters’ thesis and since then I have incorporated it into all of my subsequent research, whether it is analysing a field site or an experimental landform. In my first blog post I mentioned that I am a fluvial geomorphologist (my research is all about rivers), but since writing that post I have changed jobs and I am now using my photogrammetry skills in a whole new environment – Antarctica – working for the British Antarctic Survey in Cambridge. I have just started a new project using an archive of approximately 30,000 aerial photographs of the Antarctic Peninsula that date back to the 1940s to investigate how glaciers in this region have changed in the last 70-80 years, an area little is currently known about. This means that I am lucky enough to have access to the most amazing photography of the one of the most remote and stunning places on the planet and will be working with this for the next couple of years, which I will keep you updated on in future blog posts.

(a) Part of the archive of aerial photos held at the British Antarctic Survey, examples of historic aerial photos of the Antarctic Peninsula from (b) 1950s and (c) 1980s

Reference:

Goesele M., Snavely N., Curless B., Hoppe H. And Seitz S. 2007. Multi-View Stereo for Community Photo Collections. Proceedings of ICCV: Rio de Janerio, Brazil: 14-20 October 2007.