Author Topic: Georeference point cloud (Read 13005 times)

Gordon · « **on:** August 04, 2015, 02:37:31 AM »

Hi,

I recently ran Agisoft on a batch of images. In the reference tab I imported the GPS coordinates for each image and it seemed to recognize it and use it for processing. After building the dense cloud, the resulting points don't seem to be georeferenced as I expected. Are they supposed to be if everything was done correctly?

Thanks

Alexey Pasumansky · « **Reply #1 on:** August 04, 2015, 08:48:16 AM »

Hello Gordon,

Pleaes provide some additional details, like screenshots showing what the problem is (for example, overvew of the Reference pane and Model view).

How do you check the coordinates of the dense cloud points?

Kaja · « **Reply #2 on:** August 04, 2015, 03:41:37 PM »

I had a similar issue with my points, I needed to export a georeferenced point cloud and Agisoft kept resetting X and Y coordinates. So as a result I was getting a thin line of condensed points along Z axis. 3D mesh, orthophoto (and any other outputs) created from the same project were correct but point clouds could be exported in local coordinates only. I still haven't figured out the reason.

I'm enclosing a fragment of the point cloud in:
1) PhotoScan
2) Cloud Compare (after an export in WGS84 system)
3) Cloud Compare (after an export in local coordinates).

It wouldn't be a problem if exporting in local coordinates kept the correct values of XYZ, but these values don't look familiar at all. I should mention that the project in Agisoft PS was fully georeferenced using markers, and the coordinate system was set to WGS84.

Alexey Pasumansky · « **Reply #3 on:** August 04, 2015, 03:52:29 PM »

Hello kaja,

The second screenshot is a result of the fact that the point coordinates in WGS84 (in degrees) are treated as linear units (same for X and Y axis).

Kaja · « **Reply #4 on:** August 04, 2015, 04:59:09 PM »

Hi Alexey,

Quote from: Alexey Pasumansky on August 04, 2015, 03:52:29 PM

the point coordinates in WGS84 (in degrees) are treated as linear units (same for X and Y axis).

I'm not sure if I understand. Why are they actually treated as linear units?
Is there any way to obtain correctly georeferenced point clouds then? Would it solve the problem if I used a different coordinate system instead of WGS?

I'd appreciate your help.
Forgive me if I just started an off-topic

Kaja

Alexey Pasumansky · « **Reply #5 on:** August 04, 2015, 05:03:46 PM »

Hello Kaja,

Maybe you can specify the units in the application you are using for the point cloud visualization? What format you are actually using for the point cloud export?

Gordon · « **Reply #6 on:** August 04, 2015, 05:32:30 PM »

Quote from: Alexey Pasumansky on August 04, 2015, 08:48:16 AM

Hello Gordon,

Pleaes provide some additional details, like screenshots showing what the problem is (for example, overvew of the Reference pane and Model view).

How do you check the coordinates of the dense cloud points?

Hi Alexey,

I don't think I have any screenshots that would help. What I did before was import the GPS coordinates from a text file in the reference tab by selecting "import". I talked to someone today that said they imported from EXIF after the reconstruction was finished and then exported to a kmz file. If I add EXIF data and then do the same would you expect anything to change?

Thanks
Gordon

Kaja · « **Reply #7 on:** August 04, 2015, 05:41:10 PM »

Quote from: Alexey Pasumansky on August 04, 2015, 05:03:46 PM

Hello Kaja,

Maybe you can specify the units in the application you are using for the point cloud visualization? What format you are actually using for the point cloud export?

Hi,

I'm using a simple .txt file every time I export the point clouds.

I don't think it's a problem related to the application I use after PS processeing, because the same problem occurs in every software I tried. Besides, the points exported in local coordinates are displayed without any problem. I tried to check the units in CC, but according to their manual 'CloudCompare does not use explicit distance units'. So this should not be an issue.

What's the most surprising for me is that values along Z axis seem to be correct. Although I can't really verify if the problem is caused by XY or maybe by Z values, because I don't have a correct georeferenced point cloud to compare.

Kaja

Paulo · « **Reply #8 on:** August 04, 2015, 05:56:26 PM »

Kaja,

when you export in WGS84, X,Y values are in degrees and Z in m. It is not a cartesian reference frame as X,Y and Z have not same units or scale. In CC it takes X,Y,Z having same units and thus your problem. One degree in X or Y represents some 100 000 m !

Before exporting, you need to change your Reference system to some projected geo system like UTM and then you should get something correct...

Kaja · « **Reply #9 on:** August 04, 2015, 06:14:21 PM »

Quote from: pap1956@gmail.com on August 04, 2015, 05:56:26 PM

Kaja,

when you export in WGS84, X,Y values are in degrees and Z in m. It is not a cartesian reference frame as X,Y and Z have not same units or scale. In CC it takes X,Y,Z having same units and thus your problem. One degree in X or Y represents some 100 000 m !

Before exporting, you need to change your Reference system to some projected geo system like UTM and then you should get something correct...

Oooh, this would explain a lot! Now I see what Alexey meant by representing degrees in linear units.
I just checked it quickly and it works perfectly. Many thanks!

Forum