Laser scanner Workflow-cleaning-registration

[Topostef]

Hello all,

So since the 2.0 Version came out i have been trying to change my workflow. So up to 1.8 i would do my registering in cyclone register 360 and then save it in .e57. One thing about Register 360 is the fact that it will keep the deleted/cleaned points in the exported point cloud. Up to 1.8 this was not a major problem as you could always mask the generated panorama image and unwanted points artifacts would not show up on the final model.

Now in 2.0 I have noticed that i cannot delete/clean parts of the LSC point cloud and use it in the generate mesh with depth maps + Laser scans, since the final project will include those points. I tried Cleaning the Point cloud as well as masking the images of the cube to no avail. You can see in the attached image that the model includes the chairs that were cleaned from the point cloud. I have not yet tried importing in .pts 
because it is unstructured and i takes forever to load inside metashape.

Pointcloud cleaning is important because in any scene there are various changes between the optimal outcome and the reality of the field, like open doors moved objects etc.

Any help would be welcomed.

By the way I am also having issues with the registration as i no longer have options regarding the cameras coordinates. My workaround is to import he laser scans in a separate chunk-detect markers and save their estimated values. This way after the alignment of the scans with the images I can always import the estimated coordinates of the markers as known values.

[Dieter]

Exact the same problems here:

https://www.agisoft.com/forum/index.php?topic=15441.0

Dieter

[Topostef]

Exact the same problems here:

https://www.agisoft.com/forum/index.php?topic=15441.0

Dieter

So about the coordinates, the workaround is as follows. Since I always use coded targets for better matching between lsc and photos, I import the LSC and then detect those targets and export their estimated values and use those as known coordinates for the GCPs. This way you can transform the coordinate system through those coordinates. The other thing that works is keeping a separate chunk, only with the LSCs in their original position and then go align chunks-camera based and align the second chunk with the photos and the LSCs.

[Dieter]

Thanks for the answer.

Unfortunately we work with cloud to cloud alignement of the scans. We have to find a solution to read the coordinates of the scanners from the third-party program.


Dieter