My first scan of a flooded mine

Marie Agnes – a XVIII century silver, lead and zinc mine. Tiny gem if you look for a monument of the mining art. This spot was perfect for my first overhead 3d scan. The mine is very small – while diving as a tourist, I saw all there is just in 20 minutes. It has an interesting, irregular shape. In the flooded corridors, there are monuments of mining art, such as a wooden winch or a pipeline that once drained water, made of hollow logs. What’s important – access to the water is comfortable and easy. I just need to squeeze through the tight adit mouth on my knees and carry the equipment through about 50 m of a dry corridor.

I came prepared both in terms of diving equipment and 3d scanning plan. I had spent the last weeks in the undergrounds of the Poznań’s fortifications, testing various techniques for scanning dark, tight corridors. I wasn’t sure if the lessons I learned will apply to the flooded corridors though. I was just about to learn.

First dives went well. My idea to put the Gopro camera on a long pole that I held in front of me turned has passed the test. I carefully took the photos of the whole polace, which took me about 3 hours of shooting under water. 3 hours in such a tiny mine, where I knew one could scan a whole wreck in 25 minutes? Well, nobody said it was going to be easy. Besides the wreck is easier. If the vis is good, you can shoot from a bigger distance, covering a major part of the object. Here, in the mine I’m limited by how narrow the corrdor is. And each meter of the corridor needs to be shoot from every angle.

No air

Focusing on the task at hand is one thing, but it’s still good to stay aware of what I’m doing. Since some of the corridors are super tight, I found myself (for the first time actually) in a situation where I accidently shut my sidemount valve by rubbing against the adit walls.

Time to reconstruct

A week later I know I have a problem. I spent quite a bit of time working on the 3d reconstruction in Metashape. The shaft turned out fine. The wooden pipeline is clearly visible. The problem occurred in the corridor with the turnstile. This is the longest corridor in this adit, after passing the wooden turnstile we still have several meters of straight section ahead of us. To my despair, I have a model of the corridor, I have a model of the rest of the mine, but… separately! For some reason, the program is unable to build the short section of corridor immediately after the well, so it treats these elements as separate objects. I can’t believe it, it must be some mistake. After all, I was taking photos of all the walls, moving at a snail’s pace. It’s difficult, you have to look through all the photos manually and see for yourself. A few hours of boring work later and I already know – there are actually no photos. They end at the chamber with the turnstile and start a few meters into the corridor. Well, apparently I’m not as meticulous as I thought. I had to turn off the camera for a moment and forgot to come back to this place. There’s no need to despair. I’m packing my equipment and going to Bystrzyca.

Back to the mine

This time the diving goes very quickly. I’m prepared. I have printouts with me in which I see a three-dimensional point cloud (actually the first version of the mine map). I have marked with arrows the places where photographic data is missing. Of course, mainly the unfortunate corridor behind the turnstile, but also many smaller imperfections. I know I need to take another photo of the pipeline – there are lots of small cracks and recesses that need to be worked out. Once again, my experience with GoPro on a selfie stick is passing the test – I couldn’t reach most of these places with my hand. One hour dive and I got what I came for. Now my project consists of about 9,000 photos.

The end of the beginning

This time I’m starting from the end. First, I reconstruct the corridor behind the turnstile. Then I add photos from the turnstile area – Here it is! Where there used to be a break, now there is a beautifully reproduced rock structure. I’m adding more corridors. The program manages to find common points in about 80% of the photos – a good result considering the number of blurred photos and poor visibility in the water. Now the next step – I build a 3D mesh from the point cloud – a digital representation of the shape of the adit. After several attempts and experiments with reconstruction settings, it worked. The final stage – textures. The program must transform an image from 9,000 photos to create a graphic that, when applied to a 3D mesh, will faithfully reproduce the colors of the walls. I thought this would be the easiest process, but this is also where my lack of experience comes to light. Many photos were taken not perpendicular to the wall, but at an angle – deep into the corridor. This means that the closer part of the wall is illuminated, and the background is dark. The problem is that the program takes into account both the lit and shadowed surfaces when preparing the texture. The first approach looks dramatic – the entire mine is full of black puss. Again, some manual work is needed to tell the program which parts of the photos should be taken into account. After a dozen or so attempts, I can see the results. The colors of the rock, along with visible veins of mined metals, stand out beautifully on the three-dimensional shape of the mine. Now it’s time for some fun. I am preparing several animations showing close-ups of the most characteristic parts of the mine. I’m experimenting with 3D renders. Finally, I can show the mine in all its glory, placing the lighting wherever my imagination tells me. In real life, I wouldn’t be able to take such shots in Marie Agnes, unless after each positioning of the light I waited a few days for the sediment to settle (and the light would have to be turned on remotely anyway). Finally, I prepare a website where I collect all the materials I have created. This is how TheMarieAgnesProject.com is created – the story of how the first full scan of a flooded mine in Poland was created.

I’m glad. I’m excited, I’ve finally achieved my goal – I can show the mine as it really is. It’s impossible to be more precise. A lot of people who appreciate my work contact me – from Finland, Mexico, Thailand and, of course, from Poland. I’m happy about this, because the vision of a new project, many times bigger, is slowly becoming clear in my head. I dream of making a scan of a large, multi-level mine with corridors of min. 1 km. As I write these words, I am in the process of realizing this dream, with the help of a team of wonderful people, enthusiasts like me. But that’s a topic for a separate story.

RELATED PUBLICATIONS