SAC Segmentation of PointCloud finds 'no' lines (in simulations)

[MostafaGomaa]

Environment:
Fedora 30
Ros Melodic
Gazebo 9 Simulation

Segmentation of PointCloud into lines (in line-func.h) does not find any lines!

The segmentation calculation is used by LaserLinesThread to find laser-lines from laser point-cloud, as well as, by NavGraphGenerationThread to find obstacle lines in the map from obstacle point-cloud.

The issue could be detected on behavior level as robots repeatedly failing mps-align as it can not match laser-lines to tag-lines (since there is no laser-line data to match).

I am wondering at this point if this workes on any other configuration or in the real world (since the issue should not be related to simulation as far as I understand)

[morxa]

We haven't migrated the robots to Fedora 30 yet, so it may occur there as well.

[MostafaGomaa]

It is currently suspected that the issue has something to do with pcl1.9.
Pending on testing with pcl 1.8

[morxa]

Yes, I can confirm that the issue does not occur with pcl 1.8.

[MostafaGomaa]

An issue describing the problem has been filed
PointCloudLibrary/pcl#3439 (comment)

[vmatare]

Great, so we can just patch with 36c2bd6

[morxa]

Already in progress: https://koji.fedoraproject.org/koji/taskinfo?taskID=38524013

[morxa]

The patch indeed fixes the issue, I've created PRs for the Fedora packages:
https://src.fedoraproject.org/rpms/pcl/pull-request/2
https://src.fedoraproject.org/rpms/pcl/pull-request/3
https://src.fedoraproject.org/rpms/pcl/pull-request/4

If you can't wait for the updates, you can download my locally built RPMs here:
https://thofmann.fedorapeople.org/pcl-1.9/

[MostafaGomaa]

Cool, thanks !

[morxa]

There is now also a COPR until this has been merged in Fedora:
https://copr.fedorainfracloud.org/coprs/thofmann/pcl/

[MostafaGomaa]

This has been tested resolved!
Thanks, @morxa and @vmatare.
I mark the issue as closed!