Large format aircraft images generation DEM - DSM only created with more than 2 overlapped images

[meigeo]

Hi OpenDroneMap!
We are trying to creating DSM using ODM in docker, we are using Large format aircraft images which are around 60% overlapping for two continues aquisition. There are some areas only have two images ovelapped so they are generating wired patterns(just like I marked in the red box, the yellow boxes are image footprint)

We have found some discussion in the forum https://community.opendronemap.org/t/minimum-number-of-overlapping-images-for-point-cloud-generation/514/5?u=meigeo and changed depthmap_min_consistent_views to 2 in OpenDroneMap/SuperBuild/src/opensfm/opensfm/config.py. but we got the same result

Below I put a link include 4 sample images and EO parameters
https://drive.google.com/drive/folders/1I8mxA21QnqpFIgjTvDP6oQo8wRr5XOJ2?usp=sharing

Do you have any idea if it is possible to generate DSM with two stereo images? Or
do we need to change some other parameters?

Best regards,
Yi

[Saijin_Naib]

Welcome!

Usually that kind of smearing means that there was insufficient reconstructed points in that region and you’re seeing interpolation and gap-filling as a result.

[meigeo]

Could you please elabarate it a bit? What is reconstructed points means? is that Tie points? How could we got sufficient reconstructed points? is that possible with 2 images?

[Saijin_Naib]

Yes, they’re derived from tie points, but not every tie point gets reconstructed. I think some are filtered due to constraints like min_consistent_views and other OpenSFM settings.

You’d likely need very high overlap (80%+?) for it to fully reconstruct, and even then I would expect you to lose some points near the edges.