Dear research colleagues,
Is there a method to detect glaciers from out of space with satellite images?
Resolution is of course the bottleneck here. So coarser than 100m image data is not beneficial.
1) What Satellite Images are most beneficial for this task?
2) What type of image data might be suited best? Is a specific choice of bands helpful or do LST provide most promising results?
3) I have encountered this work below. It seems very promising, can someone evaluate?Article Google Earth Engine Open-Source Code for Land Surface Temper...
The final goal is to automatically generate masks in shape of these glaciers. For example with help of unsupervised k-means classification or if necessary supervised classification to recognize and distinguish between glaciers and everything else including snow-covered soil by color (examples attached, as you can see it does not work very well yet.). These masks are further a key element for the test part in a CNN project.
Here is a very nice project with very fine masks from lakes and sea instead of glaciers: https://github.com/JiahuiYu/generative_inpainting/issues/451
Since I work mostly with GEE, here are several LST datasets which I ask you to evaluate if you have more experience than me:
https://developers.google.com/earth-engine/datasets/tags/temperature
Cheers,
Sven Szardien
Using Supervised Classification I have achieved a better solution from Landsat 8.
As I presumed it has difficulties distinguishing between snow and glacier-ice and shadows are also problematic.
Is Land Surface Temperature somehow a solution for this? I have seen LST data from lakes that where colder than proximal glaciers.
Nasir Hameed Thanks a lot! I will try to work with the OBIA method! Sentinel-1 Images also look very promising! Incredible what is possible these days already!
Another possible approach might be to calculate slope values from SRTM's DEM and train supervisedly again. Possibly together with RGB and IR from Landsat 8 as well as data from Sentinel-1.
V. Soto I use at least resolutions of 30-60m. What satellite's rasters do you use here? Doesn't glaciers and snow have the same albedo?
Very invaluable information from you Nasir Hameed ! Thank you very very much!
Their results are incredible! According to the time I have for my actual CNN project I will see if AI-generated masks are more beneficial for the network than OBIA or pixel-generated ones.
I can update you in this discussion about my progress if wanted! You endowed me with a big leap in knowledge!
Cheers!
- Badges
- Science topic
- Similar topics
- Computer Science
- Images