I’m currently working on a project involving tall trees, taller than the houses.
I’ve got a point cloud issue. In fact, when i proceed alignment tall trees seems not to be recognize or analyze by RC. I read that it could be due to shadows or light.
Hi Hugø,
according to your images the roof should look better.
For drone dataset there is a problem with the trees. As they are captured from not so high height, there are too different between aligned images and therefor the model is not so precise. To get better trees model you need to capture also from higher height.
Could you be more specific about “not so high height”. We did 3 different try, at 40, 60 and 100 meters and we have the same issues for all test. Should we try 200 meters or more ?
40 and 60 is definitely not so high. 100 could be enough, but it also depends on the way how the images were captured.
Were you capturing using double grid? Are 190 images from all three flights?
No, it didn’t. This 190 images test has been realized between 60-80 meters (40 and 100 are from others independant test).
Also, we photographed all images in differents angles and trajectories. From top and oblique view.
I precise that we edited images in Metashape and there wasn’t any issues. Everyting is fonctionnal.
Yes, we imported images in MS then proceeded to the alignment, reconstruction and texture. It took 10 hours, that’s why we prefer using RC.
It was default reconstruction settings, we modify nothing.
but as I wrote the trees could be problematic for photogrammetry.
In this case it would be better to use all data together and also add another height. Also, capturing in cloudy days could help.
