Coregistration of ALS and TLS points clouds

[Nathalie Morin]

Hello !

I would like to test the capacity of Cloud Compare for the coregistration of 2 Airborne Laser Scanning (ALS) et Terrestrial Laser Scanning (TLS) point clouds in forest areas.
It is usually easier to identify tree tops from ALS and TLS Digital Surface Models (DSM) than from point clouds.
So my first step was to generate 2 ALS and TLS DSM with the function of the Menu Tools -> Projection -> Height grid generation, using maximum height value.
The idea is to find out what are the transformation parameters (rotation / translation) used for the coregistration of the ALS and TLS DSM, and then to apply them to the ALS and TLS point clouds, with the help of the transformation matrix (Menu Edit -> Apply transformation).

1) My first very trivial question is how can I visualize the resulting 2D image grid in (geo)tiff format in Cloud Compare ? If this is not possible, what export image format should I choose in order to be able to visualize it in Cloud Compare ?

2) Then I suppose I should use the function of the Menu Tools -> Registration -> Align or Fine registration (ICP)

Has anyone done a similar coregistration procedure of ALS and TLS data whether in image or point vector format in Cloud Compare ?

Thanks in advance !
Cheers,
Nathalie

[daniel]

I can only answer to the first question: you can't right now ;)

The Arcgrid format from ESRI will soon be integrated to CC but I'm not even sure it will help.

[Dimitri]

Hi Nathalie,

I think this should be feasible in CC. However, you do not need to export in tiff for this operation. The height projection grid will generate a point cloud in the CC left panel (sampled on a regular grid with the elevation as a scalar) which you can subsequently use for your co-registration directly in CC (as a point cloud, not a 2D raster).

I've done co-registration of raw TLS and ALS data in mountainous area with some vegetation cover with CC. It worked well by removing the vegetation first as an Iterative Closest Point calculation (as used in most surface matching techniques) does not really like the roughness induced by vegetation. So the results using only the top of the canopy could be erratic...

Best regards

Dimitri

[Nathalie Morin]

Hello Dimitri,

Thanks for the tip ! It worked fine.

1) First, I generated 2 ALS and TLS Digital Surface Models (DSM) with Tools -> Projection -> Height grid generation, based on maximum canopy height, interpolating with maximum value, leaving cells empty, in point cloud vector format.
2) Then I coregistered the ALS and TLS DSM based on on Iterative Closest Point (ICP) algorithm, using ALS as the reference, with Tools -> Registration -> Fine registration
3) Finally I applied the transformation matrix (rotation and translation parameters) to the original TLS point cloud with Edit -> Apply transformation

The only problem is that I want to do the rotation / translation in the xy plane, that is to say only around the z axis. The main reason is that I want to keep the original vegetation height difference between ALS and TLS data, which is the aim of my study. By the way, this is why there is no big interest for me in doing the coregistration of ALS and TLS using the Digital Elevation Model (DEM) based on the minimum ground points.

Here is the resulting transformation matrix :
0.999800384045 0.019821781665 -0.002483394463 -106686.671875000000
-0.019935626537 0.997974753380 -0.060405794531 23603.906250000000
0.001281018020 0.060443244874 0.998170673847 -326526.593750000000
0.000000000000 0.000000000000 0.000000000000 1.000000000000

Could anyone help me figure out how to adjust the parameters of the transformation matrix so that the rotation takes place only in the xy plane (around the z axis) and without the translation in z ? I would very much appreciate it !

Cheers,
Nathalie