3d photogrammetry modelling of cave tunnels

Techniques, equipment and issues. Also visit the NSS Survey & Cartography Section.

Moderator: Moderators

Re: 3d photogrammetry modelling of cave tunnels

Postby KariH » May 11, 2015 3:46 am

Hi, has anyone done any comparison of the coordinates/measurements that you can get from the 3D photogrammetry model and the coordinates you have taken in a more traditional way?

We are doing tests with 3D photogrammetry in underwater caves and mines , some of which are pretty deep so you don't want to spend too much time doing the measurements...

Any comments would be appreciated.

Kari
KariH
Infrequent Poster
 
Posts: 5
Joined: May 11, 2015 3:38 am
  

Re: 3d photogrammetry modelling of cave tunnels

Postby flightfollowing » Jun 8, 2015 5:36 am

Sorry for the delay. I do not yet have a quantitative comparison of accuracy between conventional survey accuracy and 3d model accuracy, but I'm working on it. Having said that, I believe 3d model accuracy to be very good based upon results of 225m long cave models compared to conventional techniques, and a few loop closures taken with 3d models wherein the loop is closed vs left open by removing a few photo frames, and evaluating the adjustment of the loop when it is completed. One will need to use a theodolite or high accuracy survey tool to evaluate in order to determine accuracy. But in the models, planar surfaces get modeled as planar, loops get closed out nicely, etc. so I think the accuracy is comparable to a quality conventional survey.
For photography, I use an Olympus tg-3 with fisheye lens set at high ISO and f/2 which does a really great job in low light, combined with a pair of zebra light 1000 lumen headlamps, no flash. You need the 120 degree flood zebralight which is the primary lighting works alone in 80% of passages, and then the auxiliary narrower beam zebralight for larger spaces in addition to the flood zebra, which gets used on full setting for almost all photos.
flightfollowing
Infrequent Poster
 
Posts: 16
Joined: Feb 2, 2015 8:48 pm
Name: Cliff Gill
NSS #: 66346
Primary Grotto Affiliation: kcag
  

Re: 3d photogrammetry modelling of cave tunnels

Postby flightfollowing » Jun 8, 2015 5:57 am

For your underwater work, you might consider using the minimum 2 stations with 1 hanging for scaling and coordinate orientation. In other words, all you must have for georeferencing and accurate geometry is 2 stations, one of which is hanging. So get an accurate distance as possible on a pair of stations, and also measure the length of the hanging portion of one of the stations making sure all three points show up well in taken photography. The two normal stations set an accurate scale for the entire model as well as good x,y orientation, and the point measured on one of those stations directly above it gives accurate z orientation. As long as you have no current in the water.
You would want to set these types of stations periodically to keep the entire model georeferenced if setting continuos conventional stations were too onerous. Once your model gets too large, (maybe much more than 1000 photos depending upon computer hardware?) it needs to be broken into segments, and each segment will need a station pair to remain georeferened.
flightfollowing
Infrequent Poster
 
Posts: 16
Joined: Feb 2, 2015 8:48 pm
Name: Cliff Gill
NSS #: 66346
Primary Grotto Affiliation: kcag
  

Re: 3d photogrammetry modelling of cave tunnels

Postby KariH » Jun 8, 2015 6:28 am

Thanks for your comments flightfollowing.

We haven't yet made any comparisons of accuracy and your suggestions sound good.

The modeling has been surprisingly successful, we made a 60min dive with a GoPro4 and two 80W LEDs attached to the nose of a underwater scooter and we could model several hundred meters of mine tunnels. Bigger halls of course caused problems.
KariH
Infrequent Poster
 
Posts: 5
Joined: May 11, 2015 3:38 am
  

Re: 3d photogrammetry modelling of cave tunnels

Postby KariH » Apr 11, 2016 5:57 am

Hi,

Here is an example of using 3D photogrammetry in cave/mine documentation.
This is a mine tunnel in an old water filled limestone mine in Finland, 58m underwater.
The model is based on frame grabs from a GoPro4 attached to an underwater scooter. The stills were processed with Agisoft PhotoScan and the animation was done in 3DS Max.

Any similar tests done somewhere?

https://www.dropbox.com/s/7s9lt2qgl4j4n ... p.mp4?dl=0

Kari
KariH
Infrequent Poster
 
Posts: 5
Joined: May 11, 2015 3:38 am
  

Re: 3d photogrammetry modelling of cave tunnels

Postby rlboyce » Apr 11, 2016 7:08 pm

IMPRESSIVE Results!!

Can you share the GoPro footage with us so we can get a before/after?

Did you take the stills at consistent frame intervals, or was there some logic to what frames you chose?
Happiness can be measured in mud/in².
User avatar
rlboyce
NSS Hall Of Fame Poster
 
Posts: 726
Joined: Jan 22, 2010 7:26 pm
Location: Chambersburg, Pennsylvania
Name: Ryan Boyce
  

Re: 3d photogrammetry modelling of cave tunnels

Postby KariH » Apr 12, 2016 12:17 am

Hi,
I usually just take 1-2 frames per second, depending on the speed I move. Some of the frames are not that sharp but it has worked quite well. With a stills camera you could of course get better quality but that would take a lot more time. I use this a lot with quite deep wreck sites and the time is a big issue.

I'll post the original video later today when I'm back home.
KariH
Infrequent Poster
 
Posts: 5
Joined: May 11, 2015 3:38 am
  

Re: 3d photogrammetry modelling of cave tunnels

Postby rlboyce » Apr 12, 2016 8:19 am

Thank you Kari. Seeing the original video would be a good data point for comparison. There is a lot of potential in the method you've demonstrated.
Happiness can be measured in mud/in².
User avatar
rlboyce
NSS Hall Of Fame Poster
 
Posts: 726
Joined: Jan 22, 2010 7:26 pm
Location: Chambersburg, Pennsylvania
Name: Ryan Boyce
  

Re: 3d photogrammetry modelling of cave tunnels

Postby KariH » Apr 12, 2016 11:17 am

Hi,

The original video can found here:
https://www.dropbox.com/s/9a9bg79bhqvsw ... 1.mp4?dl=0

It is a GoPro4 video (4K res / 25fps). From that I took 3frame grabs / second.
The Agisoft model was done with lowest quality dense cloud so a better level of detail can be produced.
The animated video is done in 3DS Max and there you can of course change the virtual camera's FOV and lighting etc etc.

Here are also some frame captures from the model in Agisoft:
(I hope they will be visible...).

All in all I would say that 3D photogrammetry is definitely worth trying!
I am currently mostly using a Sony A7S with a 14mm rectilinear lens.
GoPro 3/4 is though a very good camera to start testing with, it just needs a lot of light.

And we use A LOT of light, 2x100W LEDs on the camera and if needed a 300W and a 1000W LED carried by another diver.


Kari











Image

Image

Image

Image
KariH
Infrequent Poster
 
Posts: 5
Joined: May 11, 2015 3:38 am
  

Re: 3d photogrammetry modelling of cave tunnels

Postby rlboyce » Apr 15, 2016 6:45 pm

This is truly an excellent demonstration of current photogrammetry capability. The mesh is very close to what you see in the video, which tells me your technique is effective. The final 3D mesh is excellent. Truly an inspiration. Thanks again for sharing this!
Happiness can be measured in mud/in².
User avatar
rlboyce
NSS Hall Of Fame Poster
 
Posts: 726
Joined: Jan 22, 2010 7:26 pm
Location: Chambersburg, Pennsylvania
Name: Ryan Boyce
  

Re: 3d photogrammetry modelling of cave tunnels

Postby CaverCSE » Apr 26, 2016 4:01 pm

My first attempt at photogrammetry turned out better than expected. I created this out of 35 photos using my typical cave photography setup (2 radio triggered external flashes provided all the light). The biggest thing I've learned about doing photogrammetry is that it requires flat, boring, evenly-lit photos (which was actually pretty difficult since I'm so used to using shadows to add depth to cave photos).

Here's my map/model of the main chamber in a nearby cave (Tipton-Haynes Cave in Johnson City, Tennessee):
https://skfb.ly/NAEr
User avatar
CaverCSE
Occasional Poster
 
Posts: 35
Joined: Dec 5, 2012 11:32 am
Location: Middle Tennessee
NSS #: 49706
Primary Grotto Affiliation: Upper Cumberland Grotto and Spencer Mountain Grotto
  

Re: 3d photogrammetry modelling of cave tunnels

Postby caver.adam » Apr 27, 2016 12:46 pm

Wow, that dropbox file is seriously awesome.
User avatar
caver.adam
NSS Hall Of Fame Poster
 
Posts: 345
Joined: Feb 8, 2012 12:26 pm
Primary Grotto Affiliation: Louisville Grotto
  

Previous

Return to Survey and Cartography Forum

Who is online

Users browsing this forum: No registered users

cron