Is each circular "puddle" a scan location, there's a lot of them, if so that's a lot of 3.5 hours.
How do you tie the scan locations together. Is it a traditional total station approach with a nail in the ground or can you place some kind of reflector in the last location to tie back and chain them as you go?
Yes, each 'hole' is the location of a fixed scan location. I can adjust various settings, but the basic defaults are 0.22 degree step size and one revolution of the scan head for each step position. This results in a scan duration of about 7 minutes, after which I'll move the location and start scanning for another 7 minutes - hence why it can get boring very fast as you don't really have time to go off and do anything else, but equally there isn't much you can do other than sit back and wait. The two 3.5 hour days were a combination of Caveaton surveying and standalone scanning, in rough terms probably about half and half.
As for locating things, if we consider the Caveatron first then yes, that's total station. There are two parts to the Cavetron, a fairly standard electronic survey tool, and then in addition to this there's the LIDAR portion. If you're not interested in capturing LIDAR then the range of my laser rangefinder module is 40m, and if you having something to steady it on then you can do survey legs of that sort of range (that's how I surveyed up the hill at Snailbeach to locate the two seperate groups of data). If you DO want to capture LIDAR data, then the maximum survey leg is about 12 yds/meters/paces for reasons that will become aparent.
As I'm doing this on solo trips I have two cheap lightweight aluminium photographic light stands that I've already checked don't pull the compass. I fasten a retro-reflective target to the top of each one (about A5 paper sized). I can put one up (A), then walk about 12 paces down the level to set the other up (B). I can then take a back sight from B to A. Walk back to A and take a foresight from A to B. Each shot takes three rangefinder shots to ensure accuracy and will average/filter the compass reading to get as accurate as possible (but this is still limited to the fiddly calibration procedure you have to go through, and required me to buy a decent Suunto compass to calibrate with in the first place). I can then move A out of the way, back up a bit so there's an overlap in data and then take a 'passage' scan of LIDAR data moving towards 'B'
To capture the passage data you have to hold the red dot of the laser rangefinder steady on the target whilst you slowly walk towards it. That's a lot easier said than done! The Caveatron will ideally take a continous stream of distance and angle readings to the known survey station so it knows where in the passage the unit is, and thus where in space the LIDAR data it is capturing is located. It's all but impossible to hold it there the whole time, so internal gyro data will allow interpolation for short durations, but if that goes on for more than a few seconds you start getting warnings and must get a 'real' distance reading soon else the LIDAR scan 'fails'. You can't try bracing the Caveatron against your body as you walk, 'cos if you hold it too close then you get a scan of your boots and legs as you walk down the passage! So you're holding this thing out in front of you whilst trying to walk slow. If you take your eye off the target - such as to look where you're putting your feet - then the dot will pretty instantly fly off it. So you have to be really careful and steady. ...now imaging doing this whilst negotiating up/down a boulder slope where you can't look where you're going! ...or whilst wading in deep water where again you can't see where you're going and your hands are full so you can't steady yourself on the walls/roof! Lol. ...There's a certain knack to getting good scans! Lol.
One you've done the LIDAR scan you can walk back to recover stand A, walk past B and now set it up as survey station 'C', do a backsight towards B, walk back to B etc. etc.
If you're in a chamber than you can do a 'room' scan with the Caveatron. This means either being located at a survey station, or taking a survey shot towards a survey station and then slowly manually rotating the Caveatron as much on the spot as you can. It will use the compass data to work out where it is pointing, however it can't filter this so well when it's constantly on the move. This shows up if you rotate the unit more than once. In theory the data should stack up, but in practice you get two ghosts of the features with a rotational offset between them. Clearly one or other, or both, must contain an error in the compass data even before you consider anything metal in the area introducing distortions.
Once you have all the data you take this to the computer and use the provided processing software. This converts the raw data into XYZ data and positions each of the LIDAR point clouds in their correct positions relative to the survey stations you recorded. This XYZ data can then be opened in software such as "CloudCompare". In theory that's it, there's your 3D model. In practice you'll often find the point clouds will need an element of manual repositioning and tweaking to get them right.
With the standalone scanner that only knows the rotational angle of the scan head relative to the starting position, and nothing more. I need to manually rotate and align that point cloud data myself. To get good data you want a certain amount of overlap between point clouds anyway, so getting them aligned isn't too bad, it's just a little fiddly when you work in multiple dimensions. If I was prepared to include all the positions of the standalone scans into the stations of the survey then I could alter my processing software to take that information and apply it as offsets to the point cloud data - but in practice it would be a hell of a lot of work to accurately survey out the positions, and as the Snailbeach video demonstrates, you can pretty effectively just manually align the point clouds next to each other. Obviously it's when you come to close loops that you really find how far 'out' you are. Just need to be patient and tweak things back and forth until the data lines up.