Fossil bird: How do I create a high detail 3D model?

[Hlawktir]

Hi,

I'm trying to create a 3D-model of a bird fossil using PhotoScan. Because of a "NDA" with the scientists working on the fossil I can't upload real photos. The fossil is found on 2 plates similar to Archaeopteryx fossils: http://australianmuseum.net.au/image/Archaeopteryx-fossil-cast/

So the the whole skeleton is rather flattened: There's not a big height difference between the bones and the surrounding material - so it's rather 2- than 3-dimensional ^^
The plates are around 20 cm x 20 cm.
My aim is to produce a high-quality 3D-model with as many details as possible. What are your suggestions?

1) How important is the camera quality? Is a digicam enough?
2) What are general tips for taking the photos?
3) I'm really interested in getting high details of the small bones (length of cm to mm): Is it possible to combine photos of parts of the skeleton with photos which show the complete skeleton?
Until now I just made overview-photos, where you saw the whole skeleton: The produced 3D-model is o.k., but it lacks in details: Would it be possible to take photos of a single bone (for example the humerus or just the skull) and combine it with the complete 3D-model?
So I would like to have a model of the complete skeleton: When I zoom in, I want to be able to distinguish every single bone and also bone details (reminds me a bit of how Google Earth works).
4) How important is the masking? Does it affect the model quality or does it just help to consume less time during processing of the data?
5) I read something in the forum about taking the photos just from a 90° angle: Is this also valid for this kind of problem or is such a approach only used for orthophotos? Shouldn't I take photos from as many different angles as possible?
6) Would it be better to put the fossil on a turntable and keep the camera stationary on a tripod? Or is it better to move the camera?
7) I assume that the background shouldn't change too much. Would it help to use a bluescreen background for masking/better quality?
Is it better to take the photos with or without flashlight? If no flashlight shall be used: Should there be as much light as possible from every side or does the software need shadows (for the alignment process or to distinguish 3-dimensional features?)?
Is it possible to use PhotoScan to measure single bones' length? I'm using the Standard edition: Do i need the Professional edition?


The most important point for me is 3): I really need a high-detail model of the fossil, where you can distinguish single bones and bone details.

Sorry, if some of the questions seem stupid, but I'm relatively new to this stuff


Thanks in advance for any help!

[Wishgranter]

Hlawktir, if need can sign a NDA for this project and can help create highest possible quality of mesh and texture. No special lighting is need from description. If interested, send just one TOP image, that is clear to see how im should navigate you for taking pictures......

Best if use some DSLR camera @ shoot in RAW and highest resolution, most importaint not to shake camera.....

[Wishgranter]

1. Best with some hires DSLR, and prime lens
2. Sharp images, AWB - auto white balance. same focus
3. it is possible but its tricky sometimes..... 
And yes it is possible, when done properly you shoud get details with submilimeter precision, but need to do few tricks to it :-)
4. Masking can help getting better models ( speedier too, but mostly insignificant )
5. 90° are mostly for GIS application, for this one is best to use more angles, say for start 90° to cover area, then 60° , then 45° then 30° - need to see the model on one image. then try to get details on some "problematic" parts, that mean aditional angles, views.
6. Both solutionas have + and -, best if can rotate camera around model with proper lighting.
7. thats right, but most importaint is sharp foreground.
8. You ned proper lighting, so can use ISO 100 and proper F ( depend on camera and lens type - F8-12 mostly ) shadows do not good.
9. It is possoble, but with PRO can set proper dimensions, or use standard, preapre with some targets with known dimension  and etc.....


If need my assistancce can guide ou over email more precisely.........

3. Its possible to get better precision as with laser scanner but it need to know a lot of stuff behind it.....  This model is only one or will be more there in future ? can show good workflow..... 

[RalfH]

I agree with the points mentioned by Wishgranter.

A few additional thoughts:

The resolution of the model will depend on the resolution of the photographs. This does not necesarily mean that you need very many megapixels; 10-12 megapixels will usually be enough, you just need a large enough number of decent photographs with enough overlap.

Rather than using a turntable, I suggest approaching this similar to an aerial photography setup: take many nearly vertical, overlapping pictures (overlap 60-90%) in a lawnmower pattern, each capturing a small portion of the object (almost as small as you can get without focus problems). For example, if each image captures an area of 40 x 60 mm, horizontal movement between photographs and rows of photographs should be 5-10 mm, resulting in a total of approximately 250-500 photographs for a 200 x 200 mm object. With a 10 MPix camera you'll have a theoretical image resolution of approx. 0.02 mm. If they are well-focused and there is sufficient texture for Photoscan to work with, you should be able to get a really nice 3D model. Not necessary to take overview pictures of the whole object.

As this is a relatively small object and your camera will be quite close to it, depth of focus will become an issue. So make sure you use a macro lens (with an SLR) or macro setting (with a point and shoot) and use a relatively small aperture (larger f-stop). Near-vertical photographs will also help with depth of focus because then the object will be more or less in a plane with similar distance to the camera. Macro photography also implies that you'll probably need a tripod. Actually, because of the smaller sensor size, depth of focus will be a smaller problem with a point and shoot than with a SLR, and a good point and shoot (e.g. Canon G12) can deliver pretty much the same quality as a low-end SLR (e.g. Canon 1000D).

As mentioned by Wishgranter, low ISO (100 or even 80) and good lighting will help a lot to achieve good quality.

[Hlawktir]

Thank you a lot for your help! So rapid answers: This forum is great

I had another idea: Would it improve the output if I used some markers at the edges of the plates? E. g.: Putting a blue marker at the lower edge of one plate and a red one at the upper edge? Could this improve the alignment/quality of the model?

@Wishgranter:
I'll ask the involved people tomorrow if they accept to send you a photo. But I don't want to raise false hopes: They are really strict regarding this topic so I assume, that they won't accept sending a photo of the bird (at least for now - it depends a bit when the paper concerning the fossil is finished; sorry, can't tell you more). Would it help, if we sent you a photo of the same sediment with another fossil (not a bird, but in similar preservation conditions)?

[Wishgranter]

Hlawktir, yes, no problem with that. im can understand problematic around it.....

[3dmodelfarm]

If you have started working on your 3Dmodel,Then you can share detail with us for any queries or references.