Registering two Geo-referenced point clouds

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Registering two Geo-referenced point clouds

Zoid
Hello,

I have two point clouds obtained by Lidar and are previously Geo-referenced
as an initial alignment. You can find these clouds in the following link
from my Google drive:  

https://drive.google.com/open?id=127IiT2ntfCX3b7Rr3Y3IOo__L7-UjIYY
<http://drive.google.com/drive/u/0/folders/127IiT2ntfCX3b7Rr3Y3IOo__L7-UjIYY>  

So far I have tried: a) ICP, GICP, NDT, ICPNL with the incremental
registration as described in:
http://http://www.pointclouds.org/documentation/tutorials/pairwise_incremental_registration.php
<http://www.pointclouds.org/documentation/tutorials/pairwise_incremental_registration.php>
. Unfortunately, so far I had no success of obtaining a correct alignment
between these two maps.


Can anybody help me to properly register these two point clouds?

Thank you in advance,

Akis



--
Sent from: http://www.pcl-users.org/
_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: Registering two Geo-referenced point clouds

Sérgio Agostinho
Post an image of your point clouds overlapping with your initial
alignment guess.

All the algorithms you mentioned have a number of parameters which are
very dependent of on your data, so simply saying you tried them doesn't
tell much.

Cheers



_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users

signature.asc (836 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Registering two Geo-referenced point clouds

Zoid
Hello,

You are absolutely right on your comment. Please allow me to be a little
more specific. Let's say for the NDT and ICP algorithms I used the following
specification after down-sampling the clouds using voxel grid to a little
less than a million points:


pcl::NormalDistributionsTransform<pcl::PointXYZI, pcl::PointXYZI> ndt;
ndt.setTransformationEpsilon(0.1);

//ndt.setStepSize(0.1);
//ndt.setResolution(0.05);
ndt.setMaximumIterations(100);
ndt.setMaxCorrespondenceDistance (5);

//ndt.setEuclideanFitnessEpsilon (1);
//ndt.setRANSACOutlierRejectionThreshold (1.5);

ndt.setInputSource(cloud_origin_vox);
ndt.setInputTarget(cloud_target_vox);

ndt.align(cloud_registerd);

The initial alignment of the clouds are about X=9.5m, Y=3m, Z=0m apart
(roughly estimated). Here are some pictures of the clouds before
registration:

<http://www.pcl-users.org/file/t495484/screenshot-1521888050.png>

<http://www.pcl-users.org/file/t495484/screenshot-1521888062.png>

<http://www.pcl-users.org/file/t495484/screenshot-1521888326.png>

Any ideas?


 



--
Sent from: http://www.pcl-users.org/
_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: Registering two Geo-referenced point clouds

Antoine Rennuit-2
Is max correspondence distance set to 5 right if you say the clouds are roughly separated by 9.5m on one components?

Antoine.


-------- Message d'origine --------
De : Zoid <[hidden email]>
Date : 24/03/2018 11:56 (GMT+01:00)
Objet : Re: [PCL-users] Registering two Geo-referenced point clouds

Hello,

You are absolutely right on your comment. Please allow me to be a little
more specific. Let's say for the NDT and ICP algorithms I used the following
specification after down-sampling the clouds using voxel grid to a little
less than a million points:


pcl::NormalDistributionsTransform<pcl::PointXYZI, pcl::PointXYZI> ndt;
ndt.setTransformationEpsilon(0.1);

//ndt.setStepSize(0.1);
//ndt.setResolution(0.05);
ndt.setMaximumIterations(100);
ndt.setMaxCorrespondenceDistance (5);

//ndt.setEuclideanFitnessEpsilon (1);
//ndt.setRANSACOutlierRejectionThreshold (1.5);

ndt.setInputSource(cloud_origin_vox);
ndt.setInputTarget(cloud_target_vox);

ndt.align(cloud_registerd);

The initial alignment of the clouds are about X=9.5m, Y=3m, Z=0m apart
(roughly estimated). Here are some pictures of the clouds before
registration:

<http://www.pcl-users.org/file/t495484/screenshot-1521888050.png>

<http://www.pcl-users.org/file/t495484/screenshot-1521888062.png>

<http://www.pcl-users.org/file/t495484/screenshot-1521888326.png>

Any ideas?


 



--
Sent from: http://www.pcl-users.org/
_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users

_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users