Distorted reconstruction

Hello everyone,

We are using ODM to reconstruct pointcloud every week but sometimes we have a “problem”. Vertical face are not sufficiently densified. In order to improve this we decided to test two other flight scenario.
Usually we are flying single grid nadir but here we test the following configurations:

  • Cross pattern 5° angle, camera always looking in the flight direction, i.e. changing direction for each segment
  • Cross pattern 5° angle, camera always looking in the same direction, i.e. one time looking forward, one time looking backward

The second scenario provided us a distored pointcloud:

Do you have any idea of what have gone wrong ?
What would be your recommandations for better vertical reconstruction ?

Find the picts here:

Reference is the nadir flight

2 Likes

Are you able to make the gimbal angle more shallow? Maybe -15deg from horizon? Also, can you do 45deg or 20deg offset cross-pattern instead of 90deg?

1 Like

Can you share your logs? This looks like you got two separate reconstructions.

2 Likes

Sorry, my message was not clear.
The picture shows the two pointclouds obtained from the two reconstructions corresponding the two datasets.

Will re-run it on monday and update all output with log if it can help

3 Likes

Point cloud with cameras for the first set (not including the first image, which was blurry)

2nd set, which has 10% finer GSD

The steep face has a similar amount of holes, so images taken further off-nadir would help

1 Like

The reference point cloud shows more and larger holes in the steep face, so more off-nadir images required, as evidenced by the slight improvements in the later flights, and I think the flight path needs to be extended back out from the face a bit too, in order to get better coverage of the face and produce a more complete point cloud.

52 images 00:54:16

|Created on:|27/03/2022, 14:31:25|
|Processing Node:|node-odm-1 (auto)|
|Options:|auto-boundary: true, dem-resolution: 1, dsm: true, dtm: true, feature-quality: ultra, gps-accuracy: 8, matcher-neighbors: 10, mesh-octree-depth: 12, mesh-size: 250000, min-num-features: 12000, orthophoto-resolution: 1, pc-classify: true, pc-filter: 5, pc-geometric: true, pc-quality: ultra, pc-rectify: true, resize-to: -1, smrf-window: 12, use-3dmesh: true|
|Average GSD:|1.59 cm|
|Area:|5,233.78 m²|
|Reconstructed Points:|28,853,589|

1 Like

Thank you for your tests and hints. We will try next flight with 15° angle

Regarding the distorded aspect do you have any idea ? (If you look at my picture you see a cross section of flight 1 and flight 2 output which are not coincident)
How can we “be sure” when we design a flight plan that reconstruction will be ok ? With nadir images we never had such case.

3 Likes

Terrain following can help. Off-nadir will help. Increasing overlap as high as your platform supports will help.

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.