Hello,
I'm testing version 2.0.1, so I selected three datasets with 24mp, one with facade photos (2880 images), another with oblique and terrestrial images (3114 images) and another orthogonal/nadir (2450 images), with the objective of comparing with version 1.8.5. All datasets have already been texted and with satisfactory results in previous versions 1.7.6 and 1.8.5. The images have precise coordinates.
In version 1.8.5.
Set the alignment In workflow->Aling Photos->40k/10k, I do the gradual selection and cleaning of the sparse cloud after the alignment.
After running the optimization there was no image reset and the result is achieved without failures. The observed error was within the expected range.
In version 2.0.1.
At first, the images align automatically, I clean up the sparse cloud points, after also using the gradual selection, Projection error level 0.5, reconstruction uncertainty level 75 and Projection Accuracy 10 and take it to optimize.
In the façade image dataset, only 135 photos remain aligned. So I do the alignment manually, by groups and they align and when I repeat the selection and optimization they fix and do not lose the alignment in the project and reach the expected precision.
Ditto the problem for oblique images, where only 451 remain aligned.
In the nadir or nadir dataset with 10 degrees of inclination I did not find this difficulty after optimization.
I noticed that in version 2.0.1 there was a number of 4 or more components created, while in version 1.8.5 there was only 1 for all.
In the other stages Dense cloud, mesh, DEM, orthomosaic, I did not find any changes.
Best regards,
Topic: Alignment Error (Read 2740 times)
