The Potential of Low-Cost UAVs and Open-Source Photogrammetry Software for High-Resolution Monitoring of Alpine Glaciers
And gives xy and z accuracy estimates…
Thanks for sharing our open-access paper! I think we could show the great potential of ODM for high-resolution monitoring and mapping with our study. We hold almost 10.000 images from 10 aerial surveys, which we conducted with a self-developed low-cost UAV at the Kanderfirn Glacier in the Swiss Alps in 2017 and 2018. If anybody is interested in the images for testing ODM, please let me know. We would be happy to share them and make them publicly available. All orthophotos and DSMs from the study are freely available here.
As already discussed here and here, a more advanced georeferencing approach (e.g. higher-degree polynomials, thin plate spline) would help to further increase the accuracy of the ODM outputs. Are there any developments underway or is this more a long-term project?
I am looking forward to future developments and applications of ODM!
All the best,
Thanks for sharing all the outputs, and of course for your paper.
Regarding improvements to georeferencing, watch this branch here, which should provide some much needed enhancements to both GCP-based georeferencing as well as any georeferencing done as result of GPS info in the EXIF data.
As @smathermather said, check out the opensfm update branch, which brings significant improvements to GCPs. It would be interesting to reprocess the dataset with it and compare results.
I linked to the OpenSfM pull request, which isn’t as useful as this one which pulls most of those changes into OpenDroneMap:
Thanks for the links. It is great to see that you are currently working on this topic. I will check out the OpenSfM update branch the next days to reprocess my data sets and compare the new and old results.
I replied in the related Github pull requests (here and there), so it could too much of me talking, but I’m very satisfied.
I think the recent opensfm evolutions are very positive: I get a very good precision, not limited to 3 GCPs, and also the disappearance of the doming effect (due to camera distorsion). I suspect the latter is related since a better deformation/registration of the point cloud based on the GCPs would certainly reduce this dome shape.