Agisoft Metashape
Agisoft Metashape => General => Topic started by: itzhalpepe on May 07, 2021, 02:48:37 PM
-
Hello together,
I am producing high-resolution DSMs in a high-mountain area with snow.
If I optimized the model with some GCPs, the accuracy of the DSM increased. However, then I have some little holes in the DSM, where the confidence of the dense point is also low.
If I produce the DSM without GCPs the number of holes is clearly lower.
Make it sense to increase the number of key points(40000) and tie pojnt limit(4000) for a better cover of the holes?
It is not possible, to fill the holes with the option, because we need the z-value of the holes.
/Leon
-
Could you mind showing us your pictures of what you are doing ?
-
Hello Probert,
thanks for your reply and of course I can upload the report of this project in the attachment. Unfortunately, the size of each post is restricted by 512 kb.
Thanks
Leon
-
Hello Leon,
Do you get smaller number of holes in the resulting point cloud or DEM, if you use lower quality depth maps for the reconstruction?
-
Hello Alexey,
I didn't try it, because we need a DSM with a very high resolution and accuracy of the snow surface.
/Leon
-
Have you turn on your Satellite basemap on and off to compare it is overlap correctly ? If not, you might need an additional on some areas. I know it is hard to find. I work on 1938 aerial photos that lacks the info and I usually find a lot like rocks, roads, house ( if it is exist from the old days to today), etc..., lakes, reservoir , etc
I have run in some areas that are in snow and it is tough to find good features to match..
-
Hello Leon,
As a post processing step you could overlay the high and low quality DEMs (latter put as a bottom layer) in order to cover the gaps.
-
Hello Alexey,
yes, that could be the last possibility.
Is it possible, that the filter of depth maps in version 1.7.3 is clearly higher/more aggressive than in 1.6.5?
I have the feeling, that I generate clearly worse results with UltraCam images in 1.7.3 than in 1.6.5. Additionally, the processing time of the depth maps increased by factor 3 and of dense cloud by factor 1,5 to 2.
In the moment I am calculating the same project, which I already have with very good results- This was done with version 1.6.5. Than I can see the difference in detail.
I will send you an email, maybe you can calculate the results as well.
Probably, only my working group is working with UltraCam images, so that could be a special problem.
/Leon
-
Hello Leon,
We would probably need the sample dataset to reproduce the difference in processing time and also the logs from your side related to 1.6 and 1.7 versions.
Considerably longer processing time in 1.7 for the depth maps compared to 1.6 version could be observed on old graphic card models.