Creating a high-res digital twin of an anchient hypocaust system (Part1)

A while ago I was asked if I knew a solution to map hard to reach areas or areas that cannot be accessed by a person as crawling in could damage the existing structure. At that time, I really could not imagine what kind of structure we were talking about, but I got curious enough to get my head spinning. I asked to meet on site so that I could get a look at what kind of structure we are talking about.

Arriving on site I was baffled by what was built by the Roman Empire approx. 2000 years ago and was determined to figure out a way to document this site.

As a surveyor specializing in the field of 3D-scanning I have seen many scanning rovers at trade fairs like Intergeo, but never have I seen one hit the market. What I have seen a lot lately was “Spot the robot dog” by Boston Dynamics (thanks to social media) and I always liked the concept. But let’s be honest here, a small company can’t afford that kind of technology plus Spot is way to big for my kind of project. So, I decided to build my own Rover that would meet my requirements.

Here’s what my first concept looked like:

Not really fancy, I know... but it was the first step to building a functional unit and over the next months I did lots of research on the topic, started ordering parts and build what I now call "Nummer5".

I mainly used drone components as I figured a drone already could do all the stuff I want my rover to do, like collision avoidance, streaming a live video feed and so on. Skid steering was an issue, but I got that figured out thanks to SGS electronics. 

But getting my scanner to remote places wasn't the only issue I was facing. This obviously isn’t your everyday kind of project and I really wanted to get the maximum of quality out of it. Therefor my plan was to not only scan with my Faro, but also take high resolution images to combine the data. 

Now, I could’ve just used a Matterport or Civetta camera, as those are made for 360° photogrammetry. But being on a tight budget limited my funds and I decided to build one myself (silly me, I know). After many 3D printed builts I had a (almost) working prototype in my hands. Using 8 x 20mpx actioncams with a 170°FOV and HDR capability I was sure to capture enough imagedata at a single shot to produce quite nice image results. Or so I thought. Well, beforehand I did a lot of emailing with the support of said camera supplier and they always told me triggering 8 cameras at once wasn't a problem. As it turns out, it was a big problem after all and I ended up returning the cameras. Back to day one.

As I already own a Sony A7RII, which is my main photogrammetry camera, I started looking into Nodal Ninja and similar solutions like Gigapan. My main problem with those solutions was that I need to be able to remotely trigger them. And that’s when I found the DJI Ronin SC. 

This is looking neat, isn’t it? I thought so too. After doing some testing, my disappointment could have not been bigger. First, the Ronin can’t turn the camera around its nodal point. That’s a limitation I was willing to take IF everything else would work as expected. Long story short, it did not. The Ronin wasn’t able to trigger the camera most of the time. Setting higher focus times didn’t help and the overall performance was just mediocre. So, back it goes (big thanks to Amazon at this point).

Okay, now what? Another day another fresh start. This time I started looking in the “prosumer” section of 360° cameras. My first choice was the 200mpx XPhase Pro but I read about its fogging issue and decided not to buy it. Due to the low image resolution, I didn’t consider consumer 360 cameras like Ricoh Theta, Insta 360 One and so on. Next, I stumbled upon TECHE. The TECHE 720Pro seemed promising and while I got in touch with TECHEs reseller I got offered to test their newest 360 “prosumer” camera, the TECHE 360 anywhere. Well, I was looking for the 720 Pro, but hey it didn’t cost me anything, so why not give it a try. When the camera arrived, I was ready to get back on site. 

In my next article, I will talk about the workflow on site and the equipment used in detail. In part 3 I will talk about data processing and creating the final results.

I hope this is interesting for some of you. If you like, feel free to comment and ask questions (again, this is being split into 3 parts, so stay tuned)