Hello,
I have a UAV surveying made with a DJI M300 that successfully pass the aerotriangulation process, but at the reconstruction step it generates a 3mx mesh with holes in it.
the zone is difficult because it does not have enough visual features to extract keypoints, as you can see in the block AT result:
The question is how can I improve this keypoint extraction. I have changed the keypoints settings to high but it seems that it is not enough.
Regards,
You could try adding mesh constraint.
According to help:
Geometry constraints ContextCapture uses this data as a soft constraint in the reconstruction process: a reconstruction constraint is used only if there is no other reliable data in the area: wherever reliable 3D information can be inferred from input photos, it will override the reconstruction constraint.
Ignore this notice from help as it is no longer true - Geometry constraints are available only in ContextCapture Center Edition
Thanks for your answer, but how can I get a mesh constraint if this area was never mapped before?
You could cheat a litle bit or import the points as pointcloud from metashape.
Isn't the photos there overexposed so no automatic tie points are found?
Also have you tried reconstruction setting to fill all holes and try larger tiles so the holes are not near tile borders?
Cheat a litle bit or import the points as pointcloud from metashape.
I managed to add the geometric constraints using the ASTER DEM to Cloudcompare as Point cloud, then convert this to Mesh and simplify it to create a smooth OBJ which is then loaded to CC.
It was a pain in the *ss but it was good enough.
I hope that CC could improve the geometry using existing base layers, that could be interesting.
Add this as Idea - https://realitymodeling.ideas.aha.io/