Hello all,
I've been on the forums now for a while, and finally have my own issues to share (and solve)! This forum community is great and its nice to see that the software gets updated to reflect that. Thanks!
Let me start off with an explanation of the rig. I use Agisoft Standard 0.9.0 x64. I currently have 8 - Canon T3 cameras and 3 softboxes for lighting (60cm x 80 cm approx.) in a fixed-camera rig. I am trying to image as much of a face and head as possible. All cameras are paired up so there are 4 pairs of cameras to play with. I understand that more cameras would be better but that's what I have to work with right now. I also make effort to make sure that my lighting is as even as possible. In each configuration I have been able to always reconstruct the pairs of cameras without much issue.
I have had the whole rig in 3 major floor plan configurations. The first configuration I had was a square floor plan (each pair of cameras at a corner of the square facing in - 360 degree view of subject). The second was a pentagon (each pair of cameras at a corner with 1 corner empty). And finally a hexagon (each pair of cameras at a corner with 2 adjacent corners empty). I have had issues aligning photosets of faces and heads in all of these configurations. The system seems to be more robust as the cameras have come closer together. Has anyone found an optimal distance of placing pairs of cameras?
So all this testing (and a little failure) is fine except that I found an interesting tip on the Agisoft tips and tricks page, under aligning photos, under the question "Can I improve the alignment results for a fixed multi camera rig?" Basically using the Export Camera function I should be able to export the positions of each of the cameras in a well aligned capture. Then I can Import the positions to another photo set taken with the same camera position thus aiding in reconstruction. Someone please correct me if I have misunderstood the purpose of the import/export functions. And furthermore what are the major limitations to this process. So far I haven't been able to prove that the process works at all or even improves my data.
My process for the Import/Export technique is as follows. I made a tube about the diameter of a persons head and wrapped it with with a unique pattern (present wrapping paper). I put that in my capture volume and align all my cameras and take a picture. Using this method I can actually reconstruct the whole tube in the square floorplan (360 degree view), in a single capture photoset! I then export my cameras, because everything is aligned. I take a new photoset and then import my cameras. The "align points" dialog pops up and I select my settings. Once the camera alignment process is done some/most of my pictures are either not aligned or in the wrong place! Furthermore I haven't been able to definitively prove that my alignment is any better than if I had not used the Import Camera function. I've tried this process a bunch of times with different objects, positions, and distances with no resolution.
And so here I am. Any suggestions anyone?