Page 1 of 2

3d photogrammetry modelling of cave tunnels

PostPosted: Feb 2, 2015 9:36 pm
by flightfollowing
Here is an example of what is now feasible for inexpensive cave photogrammetry, using nothing more than camera and bright headlamp.

Carroll Cave Large Tunnel below entrance shaft - Sketchfab

Below is the original posting starting this thread:

Has anyone tried using some of the new photogrammetry and 3d modelling software to create 3d models of caves? I have briefly observed some demos of software being used to map mines and road tunnels for surveying purposes, which sounds like a it might work for cave surveying or at least documentation. I have not found much searching the web yet, but one possibility is here:

http://www.agisoft.com/forum/index.php? ... 9#msg17349

Although most of the examples above are inside and outside of a building, further down are great examples of narrow alleyways and archways. What I have found on the web are examples of one side of cave walls, which is not what I am interested in, I am interested in tunnel models.
Use of a fisheye lens is a fairly new development in support of these software applications, which I suspect is key to cave tunnel modelling.
Critical elements to perform this is a good low light camera/fisheye lens and wide angle illumination to get quality well exposed imagery, which I think I have solved at least for not too large of a tunnel.
Also, a rough 3d model may be one thing, but an accurately oriented survey quality model over much of a tunnel distance might be considerably more challenging, as the 3d models will not be well controlled as a narrow linear feature. I suspect that these 3d tunnel models will need to be controlled via conventional survey stations.

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Feb 3, 2015 9:21 am
by GroundquestMSA
Yes, it has been and is being done. The survey and cartography section of this forum has further discussion.

The advancement of cave survey is threatening its own purposes.

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Feb 3, 2015 10:08 am
by Leclused
same links as in the other topic

Here are two rreports about a 3d survey in a cave (sorry text is in dutch)

http://noumaulin.blogspot.be/2014/10/he ... in-3d.html (shows the result)
http://noumaulin.blogspot.be/2014/09/leap3d.html

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Feb 3, 2015 11:35 am
by flightfollowing
Thanks for the links. I'll try and be more specific about the suggestion, and differentiate from other threads here.

Terrestrial laser scanning/lidar:
viewtopic.php?f=26&t=7181
viewtopic.php?f=26&t=11922
viewtopic.php?f=26&t=16288
http://www.3dlasermapping.com/news-even ... pping-laun
viewtopic.php?f=26&t=16691
expensive dedicated industrial hardware, established and precise

Kinect style stereo camera/IR camera:
viewtopic.php?f=26&t=12654
experimental consumer hardware, experimental phase

Google Tango:
viewtopic.php?f=26&t=16027
experimental consumer hardware, experimental phase

Conventional simple camera photogrammetry of walls:
https://www.youtube.com/watch?v=hyCmrG3YShI
established and low cost, but not 3d enclosed tunnels?

Fisheye camera photogrammetry through cave tunnels:
this is my question. This should require no expensive or cave unfriendly equipment, just a fisheye lens camera (gopro? or tg-3 with fisheye) and good lighting, and automated postprocessing using $150 software, such as agisoft photoscan or pix4d.

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Feb 9, 2015 3:27 pm
by flightfollowing
Here are some very rough initial results of 3d models of carroll cave in Missouri, the shaft was taken on rappel into the cave, and the two other segments were walking passage acquired with the camera and headlamp mounted on the helmet, no flash, with olympus tg-3 with fisheye. Many of the frames were not well focussed or well lit, so their is lots of room for improvement. Note that the imagery was not taken with much prep at all, just walking or rappelling as normal, but with the camera firing every second and making sure the helment camera was generally pointed down cave.
These are very initial tests, I was not willing to use a nice non waterproof camera. Also, processing is very simple at these settings, and no texture was applied.

First images are the rappel down the vertical shaft:
Image

Image

and then walking passage, the blue panels are the camera stations where each photo was taken.
Image

Image

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Feb 11, 2015 1:00 pm
by rlboyce
GroundquestMSA wrote:The advancement of cave survey is threatening its own purposes.


Interesting thought. Care to elaborate?

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Feb 11, 2015 2:12 pm
by GroundquestMSA
rlboyce wrote:Interesting thought. Care to elaborate?


How about this:
flightfollowing wrote:Image

Image

and then walking passage, the blue panels are the camera stations where each photo was taken.
Image

Image


I'm joking.
Already elaborated a bunch, for what it's worth, Here.

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Mar 19, 2015 11:27 am
by flightfollowing
Here is a link to an interactive 3D model of a large section of the main tunnel in Carroll Cave, Missouri. It shows about 80 meters of tunnel below the entrance shaft. A person is in orange for scale.
Sketchfab model of Carroll Cave passage

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Mar 19, 2015 3:22 pm
by caver.adam
That's very cool. How long did it take to capture images to get 80m?

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Mar 19, 2015 4:50 pm
by rlboyce
Another question: what software did you use?

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Mar 22, 2015 11:36 am
by flightfollowing
That photography took about 15 minutes, it was somewhat more complicated than typical as it was a fairly large tunnel with two levels (most of the lower level is missing), and I ended up making two passes. Software used was agisoft photoscan. This photography process is in my opinion much much faster than sketching, and much more informative than a sketch.

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Mar 22, 2015 4:08 pm
by rlboyce
Thanks for the extra info! I agree, there are certain advantages (and disadvantages) of photogrammetry for representing a cave. Perhaps one day we'll have a solution that combines the best of both approaches. It's too bad agisoft doesn't cost a little less. Then again, it's somewhat comparable to traditional survey equipment.

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Mar 22, 2015 11:08 pm
by caver.adam
Agreed. Has some great appearance and 3D modelling. We ought to be able to figure out how to mark formations, fossils, bats, etc. in the notes so that they can be added to the map later?

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: Mar 23, 2015 12:48 pm
by flightfollowing
Reasonably large objects of interest can be viewed directly in the model, in the coarse model here, the person can obviously been seen, and reasonable sized speliothems as well. I would say that the detail available for texturing objects is inversely proportional to the amount of area to be covered, so in this instance their is a lot of cave shown in this model, so not a lot of detail can be rendered for a small feature. However, one could shoot a detailed model for any particular feature, and get LOTS of detail if not of a large area. Notes can be taken about any camera frame that has particular interest, and that camera frame can be called up in the 3d model and very precisely located along with its orientation. The photo can be viewed directly as well. For instance, if one had fish or bats to document, on could reference that info with a camera frame # to ID it in the model, or drop a marker that would be visible in the model to annotate. For inanimate objects, one could create a separate detail 3d model very easily (minutes), see attached example of a formation.
Carroll Cave Angel Formation

Also, all the survey measurements can be integrated into the model, such that the model is scaled and georeferenced. If one prefers a conventional cave map, just project the 3d model onto the XY plane and use that instead of a traditional plan view sketch, overlaying it on the line plots. Many cave features can be seen even on a 2d projection from the 3d model, so those could be manually sketched precisely, especially if one were looking at the interactive 3D model alongside it while drawing the 2d conventional map.

Note that these models were processed with fairly low detail settings, such that automated computer processing took at most a few hours for the large passage, and minutes for the formations, but using a faster computer and more time one can get much more detailed models than shown here.

Re: 3d photogrammetry modelling of cave tunnels

PostPosted: May 6, 2015 7:41 am
by caver.adam
Do you have any suggestions for people following in your footsteps on tips and tricks for getting good photographs? Did you have any specialty lighting set up? Thanks!