Agisoft Metashape 2.1.0 pre-release

[Dmitry Semyonov]

Pre-release version of Agisoft Metashape 2.1.0 is available for download.

This pre-release is considered as unstable and is not recommended for production use.

Please use the links below for download:

Agisoft Metashape Standard edition
Windows 64 bit
Mac OS X
Linux 64 bit

Agisoft Metashape Professional edition
Windows 64 bit
Mac OS X
Linux 64 bit

Agisoft Metashape Python 3 Module (requires Professional edition license)
Windows 64 bit
Mac OS X
Linux 64 bit

Agisoft Metashape Java Library (requires Professional edition license)
Download

Metashape 2.1.0 Change Log
Metashape 2.1.0 Python API Reference

Please let us know if you have any problems using this version.

Please post problems and/or questions concerning other Metashape version in separate topics.

This topic will be removed when the final 2.1.0 version is released.

Note
Project files created with 2.1.0 version can't be opened with Metashape 2.0 and earlier versions.

EDIT:
Metashape 2.1.0 build 17526 was finally released. Thank you for useful feedback.

Please use the links on the main Metashape page for download:
Download Installers

[Mak11]

Hi,

Are there any changes made to the alignment & reconstruction algorithms compared to last year ?

Mak

[Dmitry Semyonov]

Hello Mak,

The major new feature in Metashape 2.1 is geometric based alignment support for laser scans. The image based alignment didn't change from last year.

There is also a minor improvement in mesh generation algorithm in Metashape 2.1 that reduces banding in some cases, see https://www.agisoft.com/forum/index.php?topic=14934.0 for example.

[spader]

Hey Dmitry, can you elaborate somewhat on the algorithms of the geometric based alignment? Is it a traditional pairwise fine registration or is it some bundle adjustment co-aligning multiple clouds simultaneously?

[Michiel_0001]

Hi Dmitry,

We're having trouble georeferencing our imported laserscans. The scans have coordinates when exported from the laser software but once imported the coordinates have been removed. We tried to reimport the coordinates using import reference but metashape cannot find the file. Could you eleborate how this proces would work?

[Dmitry Semyonov]

Hello spader,

The approach implemented in Metashape 2.1 performs simultaneous alignment of multiple point clouds using tie points obtained with geometric feature matching algorithm (similar to bundle adjustment).

[Dmitry Semyonov]

Hello Michiel_0001,

Can you please send a support request to support@agisoft.com?

Please specify if you working with terrestrial or aerial laser scans, what steps are you performing for import and what file format are you using.

[JMR]

Hello Dmitry: Will the scan alignment take leveled status tag into account for scans with vertical axis compensation active? will prealignment or geotags in scans be taken as initials and help as it occurs with photos embeded or estimated orientations? will cloud to cloud consider the same all points in a scan regardless their range, and density around? Can you tell a bit more about the algorithms implemented?
Regards

[Dmitry Semyonov]

Hello JMR,

Currently we do not constrain point cloud transformation applied during alignment based on the leveled status tag. Thank you for suggestion for possible improvements.

Relative orientation of prealigned laser scans can be fixed if necessary by placing prealigned laser scans into a fixed group. Support for geotags as additional observations can be added in the future, but in the current version they are not supported in the user interface.

The algorithm is similar to feature based image matching, but applied to the point cloud domain. On the first step representative point arrangements are detected in the laser scans and for each arrangement a descriptor is calculated describing its local neighborhood. On the next step these arrangements are matched by comparing their descriptors. Found correspondences are used to initialize relative orientation of the laser scans and are further used as tie points during adjustment.

[Mr Miyagi]

Good Morning,

Great update!

I have bumped into 1 issue: I tried importing a number of large textured mesh models into the new version of Metashape. These work well on the last version (i have gone back and checked) Some of meshes i tested were created using other photogrammetry software's or other companies. The problem i am experiencing  is when i import a textured model the model breaks and parts of the model look like they are originating from a single location with lines (not the whole model just sections). When i switch the view to shaded, solid or wireframe the model looks perfect again.

Its only when switching to the Textured view that the model breaks. I recreated this with both OBJ, FBX and GLB file formats. We normally use UDIM texture format but ticking or unticking this box doesn't seem to resolve the issue.

I have made a sample OBJ which i can share as i just made it with open source data . it normally only happens with larger models small models all seem to work fine. It also doesn't seem to affect any models we made in agisoft though I'm not sure why.

OBJ models still have some textures though there all wrong, GLB models just go Black.

Let me know if you want the sample obj

[Alexey Pasumansky]

Hello Mr Miyagi,

It would be helpful, if you could send the sample data related to the reported problem to support@agisoft.com

If the original model has been also generated in Metashape, please also share the PDF report from the original project, where the model has been generated.

[bisenberger]

There is an issue with Build Orthomosaic >  Planar.

Here is a screenshto

Update 10/27/2023:
Oops! this post was a mistake, I didn't realize it had actually posted until now.

[bisenberger]

Hi Alexey,
Sorry for the erroneous last post  .

Past versions of Photoscan  used to include coordinates in the world file for planar projections (I have attached such a world file from 2017). The elevation was used for the "Y" axis. Could this feature be restored in this release?

Here is a screenshot of 2017 planar profiles in Global Mapper:



11/27/2023
Is it possible to install a version of Photoscan Pro (where the profile world file worked) with Metashape Pro still installed?

11/29/2023
I have posted more about this in a feature request thread, it includes a link to a 2017 Photoscan project file that has a planar orthomosaic with local coordinates.
https://www.agisoft.com/forum/index.php?topic=15956.msg69084#msg69084

[Tas]

Hello Mak,

The major new feature in Metashape 2.1 is geometric based alignment support for laser scans. The image based alignment didn't change from last year.

There is also a minor improvement in mesh generation algorithm in Metashape 2.1 that reduces banding in some cases, see https://www.agisoft.com/forum/index.php?topic=14934.0 for example.

Agisoft Team,
This is an incredibly encouraging development! Would you be willing to provide insight into whether the geometric-based alignment targets or benefits a specific workflow? Any information, even at a high level, could help make the pitch for Agisoft Metashape. For example...

• Is the geometric-based alignment intended to bolster terrestrial laser scan (TLS) alignments, specifically? Or is it also leveraged for aligning photos and laser scans?
  
• If the geometric-based alignment is leveraged to align photos to laser scans, is a particular workflow specified? E.g.,
  
  • 1) Align laser scans (or import pre-aligned/registered scans) in chunk 0, align photos in chunk 1, and align chunks with the point-based method. Maybe I'm overthinking matters, but it could be feasible to use the chunk tie point cloud to geometrically "pre-align" cameras to TLS point clouds. I'm speculating that this could be advantageous for feature detection and processing times.
    
  • 2) Or, alternatively, follow instructions for single-chunk alignment per https://agisoft.freshdesk.com/support/solutions/articles/31000168474-terrestrial-laser-scans-processing-in-metashape-2-0-0#Processing-of-terrestrial-laser-scans-with-images.
• Would you recommend any particular resources that delve into the relationship between TLS cloud geometry, depth maps, feature detection, TLS resolution/colorization, etc.? Specifically in relation to best practices for alignment, that is.
• Do you have any recommendations for comparing TLS+photo alignment in Metashape vs. other photogrammetry software? Is there an advantage or feature development focus that Metashape offers?

As always, any input or comments are very much appreciated.

[bisenberger]

Version 2.1.0 build 17262
In the Photos panel I right clicked on a photo, and chose Change Path.... In Relocate Photos I choose All images and hit OK, then navigate to the new folder and select the listed photo. When I double click an image in the Photos pane it opens the image in a tab. If I right click on an image with only one image selected and then select Check Paths it indicates No missing images detected. If I right click on an image and select Check Paths and then select All images, it indicates missing images and displays the old path. If I run Build Orthomosaic... then hit OK it indicates Missing Images and displays the old path.

11/14/2023 update
This appears due to a few missing photos in the folder. Sorry for not discovering this before the previous post. I think I am becoming senile