Advances in Image Processing Reveal Secrets from the Past

This post is also available in: עברית (Hebrew)

Historical military satellite images are now used to explore global environmental changes. Ecologists have harnessed new advances in image processing to improve analysis of declassified US military intelligence cold war photographs and detect previously unseen changes in the environment.

Deforestation in Romania and ecological damage from bombs in Vietnam are some of those changes.

The main data source for the analyses are Cold War Spy satellite images, which were collected by the US since 1960, initially to monitor the Sino-Soviet bloc. Eight satellites took pictures on film roll, which were then parachuted back into the atmosphere, where a perfectly timed US military plane snatched it mid-air.

The researchers employed drone image processing software and a rectification technique known as structure from motion in order to upgrade the filmed data.

Dr. Mihai Daniel Nita, Transilvania University of Brasov (Romania), a co-author and a pioneer of this method, commented that the structure from motion mathematical procedure behind the drone image processing software “allows us to process historical aerial or satellite images faster and more precisely than the traditional approach.” This essentially creates a historical Google-Earth style of map, for satellite imagery taken as far back as the Cold War.

“With this data we can not only map the extent of this damage with help of these images but also explore how landscapes have changed later in response to the war. For example, some of the bomb craters are now filled with water and are used as fishponds,” commented Mihai.

This work demonstrates that often our choice of baselines is dictated by data availability and that by using different data sources, we may shift the baselines against which we quantify change. The interpretation of environmental change will depend heavily on the reference points we choose.

The researchers caution that “this is a reminder to be very careful in our interpretations of environmental change. All data sources have their limitations.. A good idea is to consider integrating across multiple data sources whenever possible.”

There may be many more applications of the data, such as mapping the development of cities and built infrastructure. Future work could involve investigating ecological shock events such as war, and how this modified landscape has influenced the land-use change itself, according to phys.org.