Datasets that partially reconstruct?

The Hangar Off Topic
1

Hey everyone :wave:,

does anyone have a small dataset (less than 60 images) that does not fully reconstruct, even though it should have sufficient overlap? Examples in this category are often datasets that have vegetation and/or uniform patterns on the ground where part of the map reconstructs but some sections are left off.

I have some ideas I’d like to experiment with to improve image matching, but don’t have a good dataset to test against.

2 Likes
Improving orthophoto for tree in a forest
2

Do ground-level collects from a mobile count?

2 Likes
3

Looking for aerial only.

1 Like
4

I probably have a couple examples. Currently running through current (last processed in 2019…) and then I’ll DM you links if I see partials.

1 Like
5

I have some forest areas that won’t reconstruct at all, even tweaking the parameters accordingly. Next try will capture some off-nadir photos. Does it fit your needs?

2 Likes
6

A couple of issues affect ortho generation of forest areas.

  1. Overlap of images are reduced at canopy height.
  2. Tree tops can move and sway with even slight breezes, making match points difficult to determine.

I fly forest missions as high as possible and increase overlap (especially front) to >85%.
Off-nadir images don’t help much at all, because of the same reasons.

2 Likes