v.coregister / v.align : A module to translate and rotate a point cloud to best match one another

[stevensj]

Often LiDAR data is collected as strips (e.g. airborne), or point clouds with different sources (e.g. ground-based). When the two files are combined they might be slightly offset, resulting in nasty artefacts. It would be excellent to have a module that would allow one point cloud to be adjusted to best fit another one. The same module would also be able to pin a dataset to ground control points.

Speaking to people working with airborne lidar, I have found the misregistration to be a common problem. It can also be an issue when combining datasets from different sources e.g. LiDAR and SRTM. I think that commercial LiDAR packages for ground-based scanners may have something like this, but as far as I know there is nothing for airborne stuff.

The module may work by requiring the user to select points to match in each dataset. It could then return to parameters to give the best fit. This might be parameters to match A to B, B to A or turn both into some intermediate. It would also carry out the transformation if necessary. Maybe the parameters could be calculated on a thinned dataset to reduce processing time. A more advanced version might be able to pick the points automatically and to do some warping of the cloud, too.

[helena]

Replying to stevensj:

Often LiDAR data is collected as strips (e.g. airborne), or point clouds with different sources (e.g. ground-based). When the two files are combined they might be slightly offset, resulting in nasty artefacts. It would be excellent to have a module that would allow one point cloud to be adjusted to best fit another one. The same module would also be able to pin a dataset to ground control points.

....

We have just started to look into this - we have ground based laser scanner and we need to co-register multiple scans in 3D and we also work with a lot of airborne data (although the newest data usually come gereferenced pretty accurately). Please email me (hmitaso at unity.ncsu.edu) so that we can exchange more details about the potential module and the best methods for the task.

Helena

[benducke]

Sounds to me like a case for 3D procrustes analysis (although I have never heard of it being used for such massive datasets) -- but I guess you thought about that one already...

[neteler]

In GRASS 7, there is v.rectify which could do the job when you have 3D control points: ​http://grass.osgeo.org/grass70/manuals/html70_user/v.rectify.html

[martinl]

No activity last 4 years, closing, feel free to reopen if needed.