LIDAR Cave Scanning

ILoveCaves

Member
Hey all!
So whilst I've been working on the data sent in by quite a few of you for RoMine, I have continued to look into and utilise LIDAR, alongside photogrammetry, Here is a very early piece of my work from Jigsaw Passage (ogof daren cilau).

Long story short, the data produced can create routes that a smartphone would be able to follow, both using AR, but also localised coordinates.
Will share some more news later this year when the app is operational  :)

We really are at a new age of speleological research and recording! 
Anyone else been working on similar projects recently?

- See the Cave Video Below -
https://www.youtube.com/watch?v=smd2qp509wA&feature=emb_title
- Other Examples -
https://vala-heritage.com/?page_id=593
 

Attachments

  • screenshot.png
    screenshot.png
    8.5 MB · Views: 246

CJ

Member
gingernutcrazy said:
Anyone else been working on similar projects recently?

To a very minor extent, yes. It's very very useful for discussing dig sites in-between digging sessions (particularly to those who haven't actually seen the site). It's just a great reference to have instead of esoteric drawings/sketches, and easily exported to software for quick 3D surveys.
 

thehungrytroglobite

Well-known member
hopefully your digital routes are more accurate than your written descriptions, especially if smallcleugh is involved.
 

Attachments

  • hand-forms-l4.jpg
    hand-forms-l4.jpg
    66.3 KB · Views: 754

Rob

Well-known member
Good looking result there. Would be interested to know more about your data capture and processing. I would love to get a few Derbyshire caves scanned, but so far everything just looks WAY too much faff.
 
Rob said:
Good looking result there. Would be interested to know more about your data capture and processing. I would love to get a few Derbyshire caves scanned, but so far everything just looks WAY too much faff.

I would agree with everything Rob has said. We have messed with scans for a few years and not realy got much out of this work other than some cool looking fly though passage etc. It falls down when we want to produce a 2D map, it can be done but as Rob states it's a lot of faff. A Disto/PDA/Therion or whatever combo is much quicker and in my book gives you a better map.

The question is what do we want to use to navigate a cave. Electronic gadgets are wholly unsuitable to be used as a navigation tool in a lot of caves whereas a bit of A4 laminated paper is.

Scan data used with GiS is a different matter and this is a great tool but not much good actually underground. The BCRA Cave Surveying Group have discussed at length whether GiS is the future of cave surveying and to be honest I don't think there have been any firm conclusions made yet, but thats a whole new discussion. Cave scan data in VR is incredable and should be experienced to be believed, maybe this or more probably AR is the future.

Walk though scanning as opposed to more traditional station to station scanning may be a bit of a game changer but you still have the issues with shadows/holes in the data ect to deal with and although probably generally quicker than a Disto in walking passage it isnt going to work in your typical Yorkshire flat out one nostril in the water crawl.

I don't know if scanning is the future or not, it can be usefull and as seen above it can produce some fantastic imagery but at the moment I have given up any ideas of producing surveys with it. It is howerver, great to see the next generation of cave surveyors pushing the boundaries of technology. Great stuff, keep it up. 
 

CJ

Member
Swallowneck said:
It falls down when we want to produce a 2D map, it can be done but as Rob states it's a lot of faff. A Disto/PDA/Therion or whatever combo is much quicker and in my book gives you a better map.

Let me preface this by saying I have no experience or knowledge of surveying techniques, but is it not quite easy to view the 3D model in a top-down fashion and export this to be drawn properly as a guide? Same for side-profiles. For example, Blender can do orthographic views of the 3D capture. It's quite possible that this is a long-winded way to do something that is much easier with surveying tools (again, I have absolutely no idea and am just asking out of curiosity).
 

Rob

Well-known member
I think 3D is a valuable and engaging output in itself, in ways that 2D surveys cannot be, so I'm actually not too bothered about the 3D to 2D step. What CJ suggests of using a plan view of the 3D data to trace around may well be a reasonable option, i know of at least one person who's already trying to automate this from LiDAR data into Therion. However if it takes many many hours to generate 100m of 3D passage data alone then you might as well have just sketched it better in the cave, if all you really want is a 2D survey.

The challenge is to make the 3D data quick and faff free. I hoped that LiDAR on iPhones would do this but have seen no evidence yet. CJ, how have you found it for your example?
 
CJ said:
Swallowneck said:
It falls down when we want to produce a 2D map, it can be done but as Rob states it's a lot of faff. A Disto/PDA/Therion or whatever combo is much quicker and in my book gives you a better map.

Let me preface this by saying I have no experience or knowledge of surveying techniques, but is it not quite easy to view the 3D model in a top-down fashion and export this to be drawn properly as a guide? Same for side-profiles. For example, Blender can do orthographic views of the 3D capture. It's quite possible that this is a long-winded way to do something that is much easier with surveying tools (again, I have absolutely no idea and am just asking out of curiosity).

Yes you can do this and I have, but think passage detail, rocks,floor steps, water flow and pools etc. Top-down is just that, the view of the top of your passage. You can slice further down but what level do you slice? Blender may offer some transpareancy, I don't know, but the passage detail will not be clear. Other things such as colour also help to make a better map, how do you handle that? I'm not saying any of these points are insumountable but again we come back to the faff it takes. (at the moment)
 

CJ

Member
Rob said:
The challenge is to make the 3D data quick and faff free. I hoped that LiDAR on iPhones would do this but have seen no evidence yet. CJ, how have you found it for your example?

I find LiDAR scanning on iPhones pretty easy, but then again I'm not doing much with the data besides a mere reference. I've tried a few apps and found "polycam" to be good enough and free to export to Blender. I've only used the LiDAR for smallish chambers and spaces too, so I can't comment on the efficacy of bigger scans with a phone but I've done scans of subways (not the sandwich chain) above ground and it worked very well. I'll see if i can use that quick scan to illustrate what I'm going for. The phone can also capture areas in caves that I can't physically squeeze into, but obviously the lighting starts to become an issue since shadows are not handled well. I've heard it can really get the phone cooking too, but I've yet to experience that.


Swallowneck said:
Yes you can do this and I have, but think passage detail, rocks,floor steps, water flow and pools etc. Top-down is just that, the view of the top of your passage. You can slice further down but what level do you slice? Blender may offer some transpareancy, I don't know, but the passage detail will not be clear. Other things such as colour also help to make a better map, how do you handle that? I'm not saying any of these points are insurmountable but again we come back to the faff it takes. (at the moment)

I used to be fairly competent with Blender, but that was around 4 years ago and there have been a few UI changes since then and I recently struggled to find some features and I agree, that was a faff trying to figure it out. I dare say that these issues are well within Blenders capabilities, but it's not the easiest learning curve. Especially, if there were some "blanks" which needed filling in with my own interpretation of the mesh.

Re: cave features - if we use the LiDAR scan for a simple outline, done in a transparent, top-down, orthographic view, it should be possible to add those details in later? Again, this could be long winded since I don't know what surveying involves and how fast these things can be done, but I can breeze down passages fairly quickly with my phone and capture most things in minutes. And could easily add things in to plain 2D sketches with the scans as a reference. I need to do some scans that contain enough complexity and 2D overlap so I can play around and see what's possible. I'm hopefully out digging this week and will try to take some scans on my phone so I can have more of a play around with it.
 

Steve Clark

Well-known member
gingernutcrazy said:
Anyone else been working on similar projects recently?

We have tried various things over the last few years for underwater cave survey. 3d models built from therion data and exported through autocad dxf (quite crude). Also also some photogrammetry for underwater objects & some passages.

We have found sketchfab an excellent tool for publishing models. Allows folks to view the 3d model in a web-browser (on a half decent pc).

Some examples :

Our survey project at St Georges, France. Therion export (this does have a surface error and should be flipped n-s) :

https://sketchfab.com/3d-models/st-georges-model-surface-v01-866c76164d644b2ba8b95f766d56155e

Photogrammetry of Ginnie Springs Ballroom in Florida

https://sketchfab.com/3d-models/ginnie-springs-ballroom-v2-90f12b80ca8743a099e1a5d9e09fbae4

John also managed to build a mesh model of the first 400m of Ressel, with a single pass on a scooter with 8 radially-mounted paralenz cameras & lots of video lights. 60min dive, thousands of photos and dozens of hours of processing. I can't currently find a link to the video.
 

Steve Clark

Well-known member
The German GUE team work at the Nutlar slate mine have done some pretty impressive stuff :

https://www.youtube.com/watch?v=nQmWCl6cmpo

Sketchfab model, with annotations :

https://sketchfab.com/3d-models/nuttlar-shallow-parts-partial-3d-model-8a385a20eaa24388ae6d3c3313e2416c
 

alanw

Well-known member
This is bringing back memories. Between 1979 and 1984 I worked for Laser-Scan, a Cambridge company that specialised in using lasers to draw and digitise images. We did a lot of work for the Ordnance Survey and the Bank of England. One special project was for whoever ran the UK railways in those days: they were concerned about the state of their tunnels. They mounted a very bright light between two vertical plates at one end of a long flatbed car and a camera at the other. This projected a ring of light onto the tunnel wall. They then ran it slowly through a tunnel, taking a photograph every few seconds. The photos were passed on to us to digitise. Some of the visible bulges were very worrying. I believe the idea was to repeat the operation later and see how the digitised values had changed, but I'm not sure what happened to the project, I think it must have fizzled out.
 

CJ

Member
Swallowneck said:
You can slice further down but what level do you slice? Blender may offer some transpareancy, I don't know, but the passage detail will not be clear. Other things such as colour also help to make a better map, how do you handle that? I'm not saying any of these points are insumountable but again we come back to the faff it takes. (at the moment)

I've just been messing around with some scans and I see the issues with deciding where to slice for the purposes of a 2D survey. How is this issue solved for traditional surveying methods? Am I aiming to convey the "floor" of a passage, or to provide a general gist of the passage? Or showing the tightest constriction? e.g. an unusual hourglass shaped passage: how would that look on a top down 2D survey? My first thought would be to capture the dimensions of the ground and then provide a cross-sectional view for that part?

I've just done a very quick side profile of the dig site (and the window into the dig; which is bordered by a plank of wood, hence the incredibly straight line). I haven't provided a scale but one could be ascertained from the scans from what I understand? Half of this phreatic tube is impassable by man (at the current time) but was able to be captured. When I tried to do a top-down survey of the same site, things were a bit more challenging for the reasons you mention.


Edit: I had initially used the subway scan to illustrate survey capabilities, but everything is rectangular which makes it laughably easy to do. I suppose some nicely mined passages will be comparable, but that's not what we were talking about. 
 

Attachments

  • 4.png
    4.png
    57.9 KB · Views: 134

2xw

Active member
Rob said:
However if it takes many many hours to generate 100m of 3D passage data alone then you might as well have just sketched it better in the cave, if all you really want is a 2D survey.

Does anyone know if it still takes hours - or is that on a personal PC sans Amazon account?

Be interested to hear from OP how long the processing took and what the process was like. As well as - how nerdy does it get in terms of knowledge required?
 

nearlywhite

Active member
I worked on a very similar project with a couple of cavers in the States a few years ago that produced relatively similar results but we found that we had considerable centreline drift due to accelerometer error - have you tested this?

We only thought to test it after some disastrous loop closures  ;)
 

ILoveCaves

Member
Thanks for all the comments and help, I'm heading back this week so will try to sort a larger project out and get some more data to you all on here!

nearlywhite said:
I worked on a very similar project with a couple of cavers in the States a few years ago that produced relatively similar results but we found that we had considerable centreline drift due to accelerometer error - have you tested this?

We only thought to test it after some disastrous loop closures  ;)

Oh cool! I'd be interested in any data/info for errors and such you may have come across? Slowly making my way to see caves in Virginia....
I as of yet have not encountered any centerline drift, but with a larger survey, I wouldn't be surprised if there was a slight error with this that might need compensating for.
Swallowneck said:
CJ said:
Swallowneck said:
It falls down when we want to produce a 2D map, it can be done but as Rob states it's a lot of faff. A Disto/PDA/Therion or whatever combo is much quicker and in my book gives you a better map.

Let me preface this by saying I have no experience or knowledge of surveying techniques, but is it not quite easy to view the 3D model in a top-down fashion and export this to be drawn properly as a guide? Same for side-profiles. For example, Blender can do orthographic views of the 3D capture. It's quite possible that this is a long-winded way to do something that is much easier with surveying tools (again, I have absolutely no idea and am just asking out of curiosity).

Yes you can do this and I have, but think passage detail, rocks,floor steps, water flow and pools etc. Top-down is just that, the view of the top of your passage. You can slice further down but what level do you slice? Blender may offer some transpareancy, I don't know, but the passage detail will not be clear. Other things such as colour also help to make a better map, how do you handle that? I'm not saying any of these points are insumountable but again we come back to the faff it takes. (at the moment)

Currently, along with the LIDAR data it records both a video and takes pictures every couple of seconds, from this, Photgrammerty models can be easily produced and then compared against the LIDAR data; both offering a higher level of accuracy and error checking, but also one day hopefully allowing the likes of the smaller details like rocks and such.
Map wise a 2D and 3D model can easily be produced with the aim to add different types of layers to a scan.

Processing time-wise, it's pretty damn quick with the right CPU and a stupid amount of RAM haha!

Steve Clark said:
gingernutcrazy said:
Anyone else been working on similar projects recently?

We have tried various things over the last few years for underwater cave survey. 3d models built from therion data and exported through autocad dxf (quite crude). Also also some photogrammetry for underwater objects & some passages.

We have found sketchfab an excellent tool for publishing models. Allows folks to view the 3d model in a web-browser (on a half decent pc).

Some examples :

Our survey project at St Georges, France. Therion export (this does have a surface error and should be flipped n-s) :

https://sketchfab.com/3d-models/st-georges-model-surface-v01-866c76164d644b2ba8b95f766d56155e

Photogrammetry of Ginnie Springs Ballroom in Florida

https://sketchfab.com/3d-models/ginnie-springs-ballroom-v2-90f12b80ca8743a099e1a5d9e09fbae4

John also managed to build a mesh model of the first 400m of Ressel, with a single pass on a scooter with 8 radially-mounted paralenz cameras & lots of video lights. 60min dive, thousands of photos and dozens of hours of processing. I can't currently find a link to the video.

These are truly incredible! One of the aims would be to try to implement this into Cave diving in the future when i get around to heading that way :)

2xw said:
Rob said:
However if it takes many many hours to generate 100m of 3D passage data alone then you might as well have just sketched it better in the cave, if all you really want is a 2D survey.

Does anyone know if it still takes hours - or is that on a personal PC sans Amazon account?

Be interested to hear from OP how long the processing took and what the process was like. As well as - how nerdy does it get in terms of knowledge required?

For this one, it took a whole 15 mins to walk and slowly scan, and about 15 mins of processing and that's on my laptop! The process can definitely be streamlined and I'll let you know once I'm home and can load my Mine scans into the PC.

 

nearlywhite

Active member
From memory the drift was equivalent to almost 3% a station which would make it massively fall short of a BCRA grade 5 survey. It's quite a large error.

There was an idea floating around 10 years ago to have temporary brightly coloured reference stations which you could then accurately centerline, then using a similar warping as is done in Therion to correct for error.

Not caved in Virginia but have enough contacts if you need any. Have many more contacts for the better caves in Tennessee
 

nobrotson

Active member
I assume you're using an iPhone 12 pro or iPad pro? Would be good to know your exact hardware-software setup.

What density of points can you achieve? what is the resolution of the scan?

I can think of a lot of very useful geological applications of this in a cave, particularly regarding looking at passage geometry along fractures for example. However as others have stated I'm not sure its that useful for producing traditional cave maps. Interesting to hear about Rostams error sources.

For cave surveying using VR, check out TunnelVR
 
Top