Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Already on GitHub? Sign in to your account

Implement error to DEM for pc_align #109

oleg-alexandrov opened this Issue Feb 28, 2014 · 1 comment


None yet
1 participant

oleg-alexandrov commented Feb 28, 2014

For pc_align, when the reference point cloud is an actual DEM, a more useful error metric may be not distance from a source point to the closest reference point in the DEM cloud, but rather the distance to the closest point on the interpolated DEM surface. The idea here is that a DEM is not just a collection a points, it is in fact a continuous surface which happens to be sampled on a given grid. Currently, if the grid of the DEM is coarse, a source point may be very close and well-aligned to the DEM surface, yet far from any actual grid samples in the DEM, making the alignment error look worse than what it already is.

I already cooked up something like this at David's request as a separate tool, and there one just projected straight down from the source point to the DEM, which is an approximation (overestimate) to the true distance to the DEM surface but better than what is now in pc_algin.

This should be an option, like --compute-errors-to-dem-surface. The alignment algorithm itself will still use, as now, just the DEM values on the grid, treating it as a point cloud, so the new thing would be just statistics for reporting purposes before and after alignment.


oleg-alexandrov commented Aug 31, 2015

This functionality has been merged into pc_align and is the default. The old approach is still there and invoked via a documented switch.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment