"Camera Stations"

[ekbmuts]

I shot a china doll under perfect lighting conditions (strobe) and from a number of static camera positions.  The first position was head-on to the model, the next slightly angled down, the next slightly angled up, etc.  I did this to ensure that I captured all undercuts in the doll and the top of her head.

The doll herself was standing on a circular scale sheet with various markings so that I can indicate scale in Photoscan.

I rotated the doll by hand; she was not on a turntable and so her exact center may have varied somewhat but this would have been very minimal.

I took about 15-20 shots for each revolution.  The shots were taken at ISO 200, 1/125, f11 and are nice and crisp.

It's my understanding that these photos need to be separated into different camera groups within the same chunk once imported into Photoscan and that these individual camera groups need to be named "Camera Station."  I read this in the manual.

So far so good I think.

A few questions:

1.  Is this part of the workflow accurate?  In other words, does anyone see anything wrong with my sequence?

2.  Can I name these camera groups "Camera Station #1", "Camera Station #2", etc.  Or do they all have to be called just "Camera Station"?

I'm posting this because I'm not getting the best results.  The dense cloud is coming out flat and with no depth which mystifies me because this is a project being done under controlled and perfect conditions.  But my results are way, way better when I go out in the field under a cloudy sky and shoot a wall hand-held (for example).

Thanks in advance.

Jon

[macsurveyr]

In the scenario you describe, the cameras should NOT be grouped. Every camera location is in fact unique, in relation to the subject, as if you moved around the doll.

If you moved the doll independent of the circular scale sheet, that will also be an issue. It would be best to have the circular scale sheet and the doll on a turntable and move both together so that the scale and the doll will be modeled together in the same relative 3D space. Any background that does not move with the doll should be masked out as well.

Hope that is clear.

Tom

[ekbmuts]

Tom,  Thanks for the quick response.  I see what you're saying.

On the scale sheet, right - it certainly didn't stay tracking with the model.  In fact, the sheet stayed fairly still while the model rotated on top of it.  I get it  - I'm giving Photoscan a moving target when I don't move them together.  Unless of course I mask out the scale sheet completely before aligning.

In my final "bracket" of shots where I shot from a plane beneath the models base, I elevated the model by about 3" so that I could shoot all the undercuts.  Disabling these cameras has made a world of difference in my results. Just disabling this group of photos increased my Tie Points from 857 to 1,117, which is a nice improvement.  I don't quite understand why this is though.  It's not like the model changed in any way, just her up and down position (y).  Maybe because the scale sheet is not present in these shots?

In any event, I think, based on your advice, that I'll reshoot her properly and see how she comes out.

Thanks again.  Very appreciated.

Jon