drone to laser scan registration

Please post all open source software related items here, eg MeshLab
User avatar
danielgm
V.I.P Member
V.I.P Member
Posts: 155
Joined: Thu Jun 20, 2013 10:45 am
Full Name: Daniel Girardeau-Montaut
Company Details: CloudCompare
Company Position Title: Administrator
Country: France
Linkedin Profile: Yes
Location: Grenoble, France
Contact:

Re: drone to laser scan registration

Post by danielgm » Tue Jan 06, 2015 3:37 pm

If the scale is right, then the clouds may have non negligible dissimilarities that prevent the ICP algorithm from converging to the right position: is the coverage of the reference/model cloud is the same (or wider) than the aligned/data cloud? Otherwise you can also try to check the 'Enable farthest point removal' (so that CC ignores those differences).

Eventually, don't hesitate to post (or send me at cloudcompare [at] danielgm.net) some snapshots of both clouds.
Daniel
CloudCompare admin
http://www.cloudcompare.org/

User avatar
blazaj
V.I.P Member
V.I.P Member
Posts: 465
Joined: Tue Jan 26, 2010 1:07 pm
Full Name: Jiri Blazek
Company Details: BW precision systems sro
Company Position Title: Managing Director
Country: Czech Republic
Linkedin Profile: No
Contact:

Re: drone to laser scan registration

Post by blazaj » Tue Jan 06, 2015 4:53 pm

You can try to do rough manual registration to see if the scale and registration of individual pointclouds is alright. this can be done in Scene, suppose Cyclone can do something similar.

tbwester
V.I.P Member
V.I.P Member
Posts: 886
Joined: Fri Apr 20, 2012 1:49 pm
Full Name: Thad Wester
Company Details: WeWork
Company Position Title: Head of Reality Capture
Country: USA
Linkedin Profile: Yes
Location: SC/NYC

Re: drone to laser scan registration

Post by tbwester » Tue Jan 06, 2015 6:17 pm

blazaj wrote:You can try to do rough manual registration to see if the scale and registration of individual pointclouds is alright. this can be done in Scene, suppose Cyclone can do something similar.

I have done that.

I checked adjust scale, remove outliers, 100,000 samples. 50 iterations. See image for the result. (drone in color, laserscan in red)

I'm starting to wonder if the drone point cloud is just off by that much. I can't seem to get it aligned better than 1 foot. The detail in the surface is really good, resolving features < 1 foot, but when comparing to the laser scan, the data seems to drift.

I cant tell if it is an error in registration, or if its just the drone being inaccurate globally.
You do not have the required permissions to view the files attached to this post.

tbwester
V.I.P Member
V.I.P Member
Posts: 886
Joined: Fri Apr 20, 2012 1:49 pm
Full Name: Thad Wester
Company Details: WeWork
Company Position Title: Head of Reality Capture
Country: USA
Linkedin Profile: Yes
Location: SC/NYC

Re: drone to laser scan registration

Post by tbwester » Tue Jan 06, 2015 7:06 pm

Ok. I was able to get a decent result when I left the scale alone, and checked enable farthest point removal. Even so, the point cloud was off by about 1.5 feet consistently. So I moved it together manually and it looks pretty good (see attached). Some areas are off by 2 feet and others are within .4 feet.

So, its not very accurate. Even though it looks accurate by resolving small features - the global site accuracy is poor. Keep in mind this is a gopro.
You do not have the required permissions to view the files attached to this post.

User avatar
blazaj
V.I.P Member
V.I.P Member
Posts: 465
Joined: Tue Jan 26, 2010 1:07 pm
Full Name: Jiri Blazek
Company Details: BW precision systems sro
Company Position Title: Managing Director
Country: Czech Republic
Linkedin Profile: No
Contact:

Re: drone to laser scan registration

Post by blazaj » Wed Jan 07, 2015 7:27 am

I'd say good results with gopro. It has wide angle lens and not the best quality for photogrammetry. Curious what would you get with better camera

User avatar
danielgadowski
V.I.P Member
V.I.P Member
Posts: 451
Joined: Mon Jun 30, 2008 3:49 pm
Full Name: Daniel Gadowski
Company Details: London Underground
Company Position Title: Laser Scanning Co-Ordinator
Country: UK
Linkedin Profile: Yes

Re: drone to laser scan registration

Post by danielgadowski » Wed Jan 07, 2015 10:10 am

I agree with Blazaj.
Gopro is not a platform for this type of aplication. It has too big distortions. If you want to have good results you should use a fidderent cammera, preferably a DSLR with calibrated lense. Also you could use some sort of photogrammetric targets (ping pong balls, small round plastic plates or something like that) to tie the images better together and give yourself accurate scale if you measure them.
Good luck with your endevours :)
Dan

tbwester
V.I.P Member
V.I.P Member
Posts: 886
Joined: Fri Apr 20, 2012 1:49 pm
Full Name: Thad Wester
Company Details: WeWork
Company Position Title: Head of Reality Capture
Country: USA
Linkedin Profile: Yes
Location: SC/NYC

Re: drone to laser scan registration

Post by tbwester » Wed Jan 07, 2015 2:44 pm

Of course, but my copter will only fit a gopro:) It is merely an experiment.

I correct the photos to remove the distortion using PTlens.

Even with the shoddy registration and gopro. My result was within 4% of the laser scan volume.

User avatar
Matt Young
Global Moderator
Global Moderator
Posts: 3830
Joined: Fri Nov 30, 2007 3:03 pm
Full Name: Matt Young
Company Details: Baker Hicks
Company Position Title: CAD-BIM Lead
Country: UK
Linkedin Profile: No
Has thanked: 6 times
Been thanked: 3 times

Re: drone to laser scan registration

Post by Matt Young » Wed Jan 07, 2015 2:52 pm

For an un-level area with anomalies I think 4% is probably ok? On a level surface you would want to aim closer to 2% tolerance I reckon, I'm no expert. Pretty good scanning results from a drone anyway!
If you want to make something simple, then take one thing away that still allows you to do the job. Don't add two more things, and say they both make it easier.

tbwester
V.I.P Member
V.I.P Member
Posts: 886
Joined: Fri Apr 20, 2012 1:49 pm
Full Name: Thad Wester
Company Details: WeWork
Company Position Title: Head of Reality Capture
Country: USA
Linkedin Profile: Yes
Location: SC/NYC

Re: drone to laser scan registration

Post by tbwester » Wed Jan 07, 2015 3:10 pm

Matt Young wrote:For an un-level area with anomalies I think 4% is probably ok? On a level surface you would want to aim closer to 2% tolerance I reckon, I'm no expert. Pretty good scanning results from a drone anyway!
Thanks Matt that is good to know. Not the right tool for every project, but it is a tool.

User avatar
Matt Young
Global Moderator
Global Moderator
Posts: 3830
Joined: Fri Nov 30, 2007 3:03 pm
Full Name: Matt Young
Company Details: Baker Hicks
Company Position Title: CAD-BIM Lead
Country: UK
Linkedin Profile: No
Has thanked: 6 times
Been thanked: 3 times

Re: drone to laser scan registration

Post by Matt Young » Wed Jan 07, 2015 3:31 pm

Drones will only get better I think. Ten years ago I would of laughed at the idea of scanning from a radio control drone, but now it's more than a reality :)
If you want to make something simple, then take one thing away that still allows you to do the job. Don't add two more things, and say they both make it easier.

Post Reply

Return to “Open Source Software”