I already posted this once, but it isn’t showing up in my contributions or in the discussions so I’m assuming it failed to post? It may have been removed for some reason (I don’t think it broke any rules) but I never got a message telling my that it had been or why so I’m just gonna repost it.
So I’ve been playing around with Reality Capture for a while now, but I’m still not sure what the majority of the settings do or how they work, so just imagine I’m completely new and have no idea what I’m doing (that second part is still pretty much true).
So I have a camera rig built for capturing full body scans of people, and I’m having trouble getting consistent results. My current workflow is about as basic as it gets; I import my photos and align them, then I set the reconstruction region and ground plane manually. I then calculate the model, check the topology, clean the model, and close the holes. After that I simplify it to 150,000 tris, color it, and texture it.
This works kind of okay but a lot of the time things get split into multiple components during alignment, and the model quality isn’t very good. After seeing the model another team produced using a camera array similar to mine it made me a bit self conscious and I decided I need to do better.
One glaring problem I’m having is that I need all this to be automated, and that means I can’t be setting the recon region and ground plane manually. The only way I can reliably import those would be if I saved the camera positions, and so far that’s resulted alignment’s with multiple components unless I export them as exact, which completely butchers my model 90% of the time.
I assume this is caused by a few factors. One, my rig is made out of bendy aluminum so the camera positions are always slightly shifting around, and two, I’m relying purely on the camera positions given by previous scans that did get everything into one component, so the saved positions aren’t very precise.
I’ve read a bit about flight logs, and they sound pretty much perfect for my needs, but I can’t figure out how to make one. There’s a button for importing flight logs, but as far as I can tell there isn’t one for exporting them. Also I know you can use gps coordinates to get precise camera positions but I have no clue how.
I see a lot of full body scanners have mats that you stand on with qr codes on them. I understand why they’re on the mats (I think), so they can cut off the bottoms of their feet more easily, but I’m not sure why they have qr codes. I assume it’s for control points, which I should probably be using, but the problem there is that I’m gonna be disassembling and reassembling the array a lot in varying environments. I feel like that would cause problems as 1) It’s gonna be hard to get a mat in the same place every time, and 2) the cameras are only attached to the frame with mounts, they aren’t fixed to it, so they might get bumped around during transport.
I’d also much rather not have to redo the control points every time I set it up, especially since it would make it more difficult to automate.
One last thing, I need the scale on all the models to be pretty much the same as they are in real life but I always get confused when I start reading about setting up coordinate systems and whatnot. Honestly I don’t even know what euclidean coordinates are.
So what can I do to improve my workflow, and how would you suggest I get over some of these technical hurdles?
TL;DR: The models coming out of my full body scanning rig suck, pls help.
-JS