Computer now recognizes avalanches on satellite images

Date:

Avalanche monitoring made easier thanks to technology: researchers at the Institute for Snow and Avalanche Research (SLF) in Davos have developed a method that can automatically detect avalanches and map them on satellite images.

The SLF announced that satellite images can be used to map the entire Alpine region on a large scale. This bird’s eye view is particularly revealing. Because it gives an answer to the areas where avalanches occur repeatedly, how often this happens and what route the avalanches follow.

Until now, mapping has been laborious manual work: you have to search the images on the screen for avalanches by eye and manually trace their outline, as the SLF wrote.

PhD student Elisabeth Hafner and her colleagues therefore developed an image analysis method based on machine learning. To do this, Hafner first mapped more than 24,000 avalanches based on satellite images from January 24, 2018 and January 16, 2019.

First results “unsatisfactory”
This manual mapping formed the basis for training the computer – teaching him what is an avalanche in the shot and what is not. In addition to the optical data, topographical information was also used. Because avalanches almost only occur on slopes steeper than 30 degrees.

According to Hafner, the initial results were “not very satisfactory”. However, the researchers suspected that this could be because even experts don’t always agree on the existence and shape of avalanches.

Hafner then had a satellite image manually mapped by five avalanche experts and compared their agreement. The agreement was found to be higher in sunny, high-visibility areas than in shady areas.

In general, the experts recognized the same amount of avalanche area as the model. This shows that the computer recognizes avalanches almost as reliably as different experts.

In a next step, Hafner will investigate how experts work with manual mapping and why they judge the recordings differently in certain cases. “If we identify these uncertainties, we have a chance to adjust our model accordingly,” the doctoral student said in the statement.

Source: Krone

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Share post:

Subscribe

Popular

More like this
Related

Ancelotti complains about the schedule against Mallorca

“The minimum is to rest for 72 hours, the...

Manhunt – Unknown robs 94 year old car and flees

The key of a 94-year-old man was violently snatched...