Can someone test a program with point clouds?

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
10 messages Options
Reply | Threaded
Open this post in threaded view
|

Can someone test a program with point clouds?

Pierre Abbat
I'd like to release PerfectTIN 0.4.0, but I need someone to verify that the
output is free of spikes. The person who was going to do this cannot meet his
coworkers because of the pandemic. If you work with point clouds (preferably
in LAS format, but it can read PLY if compiled with a library) and can view a
TIN and turn it around, please let me know.

Pierre
--
La sal en el mar es más que en la sangre.
Le sel dans la mer est plus que dans le sang.



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Rajat Shinde
Hi Pierre,

I would be very happy to do it. My PhD thesis involves LiDAR Point Cloud processing and these days I am fully covered up with LAS/LAZ files. Though, I have not used PerfectTIN till now, but I can see the earlier releases available at https://github.com/OSGeo/perfecttin/releases.

Please suggest me on how to proceed. 

Kind regards,
Rajat

On Sat, Apr 18, 2020 at 12:11 AM Pierre Abbat <[hidden email]> wrote:
I'd like to release PerfectTIN 0.4.0, but I need someone to verify that the
output is free of spikes. The person who was going to do this cannot meet his
coworkers because of the pandemic. If you work with point clouds (preferably
in LAS format, but it can read PLY if compiled with a library) and can view a
TIN and turn it around, please let me know.

Pierre
--
La sal en el mar es más que en la sangre.
Le sel dans la mer est plus que dans le sang.



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss

_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Pierre Abbat
On Friday, April 17, 2020 2:57:13 PM EDT Rajat Shinde wrote:
> Hi Pierre,
>
> I would be very happy to do it. My PhD thesis involves LiDAR Point Cloud
> processing and these days I am fully covered up with LAS/LAZ files. Though,
> I have not used PerfectTIN till now, but I can see the earlier releases
> available at https://github.com/OSGeo/perfecttin/releases.
>
> Please suggest me on how to proceed.

Thanks! Here are the steps:

Clone the repo (the latest commit should be on 2020-04-07 or later).

Check out 0.3.3 and run it on some point clouds, looking for some where the
TIN has big spikes (most likely in holes).

Check out later commits and run them on the same clouds. The spikes should
become smaller in horizontal area (they may be less than a millimeter thick)
and then disappear. Thin spikes may appear in places they weren't before (I've
seen them in falsely rough asphalt that is an artifact of photogrammetric
processing of pine needle shadows), but should disappear by the latest commit.

Also, would you like to contribute a translation?

Pierre
--
gau do li'i co'e kei do



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

adam steer-2
In reply to this post by Pierre Abbat
Hi Pierre

I am pretty point cloud heavy also. I’ll see if I can spin up perfectTIN and try it out!

Regards

Adam

On Sat, 18 Apr 2020 at 04:41, Pierre Abbat <[hidden email]> wrote:
I'd like to release PerfectTIN 0.4.0, but I need someone to verify that the
output is free of spikes. The person who was going to do this cannot meet his
coworkers because of the pandemic. If you work with point clouds (preferably
in LAS format, but it can read PLY if compiled with a library) and can view a
TIN and turn it around, please let me know.

Pierre
--
La sal en el mar es más que en la sangre.
Le sel dans la mer est plus que dans le sang.



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss


--

_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Rajat Shinde
In reply to this post by Pierre Abbat
-- Resending it again with reduced size, please ignore if received twice. --

Thanks for the detailed instructions.
Initially, I tried with a huge point cloud scene but it was getting difficult to visualize so then I sampled a small subset out of it. The subset is having a valley sort of and hence it was expected to produce steep triangles. All the experimental results are in the attached zip file. The experiments were done for Release 0.3.30.3.6 and 0.4.0rc1 (master).

As mentioned, the spikes are visible in 0.3.3 but it gets pretty smooth and smaller in 0.3.6 and 0.4.0rc1. Please check it. I have also added some screenshots for visual comparison.

Not sure about the attachment size on the mailing list, so the results can also be accessed here.

I hope of not missing out on anything. Please let me know if I did.
I would love to see Dr Steer's reply; he is really point cloud heavy and I have been following his articles lately. :)

Kind regards,
Rajat

On Sat, Apr 18, 2020 at 3:37 PM Rajat Shinde <[hidden email]> wrote:
Thanks for the detailed instructions.
Initially, I tried with a huge point cloud scene but it was getting difficult to visualize so then I sampled a small subset out of it. The subset is having a valley sort of and hence it was expected to produce steep triangles. All the experimental results are in the attached zip file. The experiments were done for Release 0.3.3, 0.3.6 and 0.4.0rc1 (master).

As mentioned, the spikes are visible in 0.3.3 but it gets pretty smooth and smaller in 0.3.6 and 0.4.0rc1. Please check it. I have also added some screenshots for visual comparison.

Not sure about the attachment size on the mailing list, so the results can also be accessed here.

I hope of not missing out on anything. Please let me know if I did.
I would love to see Dr Steer's reply; he is really point cloud heavy and I have been following his articles lately. :)

Kind regards,
Rajat

On Sat, Apr 18, 2020 at 1:57 AM Pierre Abbat <[hidden email]> wrote:
On Friday, April 17, 2020 2:57:13 PM EDT Rajat Shinde wrote:
> Hi Pierre,
>
> I would be very happy to do it. My PhD thesis involves LiDAR Point Cloud
> processing and these days I am fully covered up with LAS/LAZ files. Though,
> I have not used PerfectTIN till now, but I can see the earlier releases
> available at https://github.com/OSGeo/perfecttin/releases.
>
> Please suggest me on how to proceed.

Thanks! Here are the steps:

Clone the repo (the latest commit should be on 2020-04-07 or later).

Check out 0.3.3 and run it on some point clouds, looking for some where the
TIN has big spikes (most likely in holes).

Check out later commits and run them on the same clouds. The spikes should
become smaller in horizontal area (they may be less than a millimeter thick)
and then disappear. Thin spikes may appear in places they weren't before (I've
seen them in falsely rough asphalt that is an artifact of photogrammetric
processing of pine needle shadows), but should disappear by the latest commit.

Also, would you like to contribute a translation?

Pierre
--
gau do li'i co'e kei do



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss

_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Pierre Abbat
On Saturday, April 18, 2020 6:11:24 AM EDT Rajat Shinde wrote:
> -- Resending it again with reduced size, please ignore if received twice. --

Reduced size??

> Thanks for the detailed instructions.
> Initially, I tried with a huge point cloud scene but it was getting
> difficult to visualize so then I sampled a small subset out of it. The
> subset is having a valley sort of and hence it was expected to produce
> steep triangles. All the experimental results are in the attached zip file.
> The experiments were done for Release 0.3.3
> <https://github.com/phma/perfecttin/releases/tag/0.3.3>, 0.3.6
> <https://github.com/phma/perfecttin/releases/tag/0.3.6> and 0.4.0rc1
> (master).
>
> As mentioned, the spikes are visible in 0.3.3 but it gets pretty smooth and
> smaller in 0.3.6 and 0.4.0rc1. Please check it. I have also added some
> screenshots for visual comparison.
>
> Not sure about the attachment size on the mailing list, so the results can
> also be accessed here
> <https://drive.google.com/drive/folders/1D9fAAooisQCO5BH_Ir12pyLe84Cqmy0z?us
> p=sharing> .
>
> I hope of not missing out on anything. Please let me know if I did.
> I would love to see Dr Steer's reply; he is really point cloud heavy and I
> have been following his articles lately. :)

I see some screenshots of PerfectTIN and some files which I won't try to
download until 2:00, so that I don't run down my data allotment. What tool did
you use to turn the TINs so that you could see spikes? Do you have some point
clouds with holes where objects such as houses or cars were removed?

Pierre
--
ve ka'a ro klaji la .romas. se jmaji



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Pierre Abbat
In reply to this post by Rajat Shinde
The point cloud has only 44367 dots. The cloud I use for quick tests (and
which did produce spikes, because some cars were cut out of it) is 250024
dots. Others I've worked with have 13 million dots or hundreds of millions.

I opened the .ptin file from 0.4.0rc1 with SiteCheck. It looks like rough
terrain with nowhere near enough dots to define it within a decimeter. Or maybe
there are lots of trees in the point cloud.

You may want to run the program on the whole point cloud with looser
tolerance.

Pierre

--
I believe in Yellow when I'm in Sweden and in Black when I'm in Wales.



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Rajat Shinde
Hi Pierre, Thanks!

Okay. I mis-understood "turning the tin" earlier and also interpreted the generated triangles as spikes. Now, I got it. 

I am not having any point cloud file with cars cut out of it but I am positive of some point clouds with buildings cut out. I would try running the program again on such complete scene (with looser tolerance) and get back to you with results.

Kind regards,
Rajat

On Sun, Apr 19, 2020 at 1:39 PM Pierre Abbat <[hidden email]> wrote:
The point cloud has only 44367 dots. The cloud I use for quick tests (and
which did produce spikes, because some cars were cut out of it) is 250024
dots. Others I've worked with have 13 million dots or hundreds of millions.

I opened the .ptin file from 0.4.0rc1 with SiteCheck. It looks like rough
terrain with nowhere near enough dots to define it within a decimeter. Or maybe
there are lots of trees in the point cloud.

You may want to run the program on the whole point cloud with looser
tolerance.

Pierre

--
I believe in Yellow when I'm in Sweden and in Black when I'm in Wales.



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss

_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Pierre Abbat
On Monday, April 20, 2020 2:16:20 AM EDT Rajat Shinde wrote:
> Hi Pierre, Thanks!
>
> Okay. I mis-understood "turning the tin" earlier and also interpreted the
> generated triangles as spikes. Now, I got it.
>
> I am not having any point cloud file with cars cut out of it but I am
> positive of some point clouds with buildings cut out. I would try running
> the program again on such complete scene (with looser tolerance) and get
> back to you with results.

Okay, let me know what you find.

Dr. Steer, how are you doing with the program?

Pierre
--
sei do'anai mi'a djuno puze'e noroi nalselganse srera



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss
Reply | Threaded
Open this post in threaded view
|

Re: Can someone test a program with point clouds?

Pierre Abbat
I'm in Georgia visiting another surveyor, who asked me to write the program to
begin with, and got it running on his NUC in such a way that he can try
different versions if I tell him the commands to run. It's running on a full-
size (56 million) point cloud, which may be too big for this NUC's RAM. I've
found a few problems:

* The NUC has Wayland. I've never run it on Wayland before, the pulldown menus
are lower than they should be, it outputs Wayland errors, and when I hit a key
to stop the screen saver, PerfectTIN crashed. "The Wayland connection broke.
Did the Wayland compositor die?"

* When I tried to restart it from the checkpoint file, it said it was corrupt,
and some menu items remained grayed out even when I cleared it.

* It got stuck for a long time with eleven big triangles all sharing a common
corner. This is an old problem, I know what's going on, and I'm going to add a
button as a workaround.

Pierre
--
li ze te'a ci vu'u ci bi'e te'a mu du
li ci su'i ze te'a mu bi'e vu'u ci



_______________________________________________
Discuss mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/discuss