Very small (13mm x 8mm x 5 mm) objects: problem with tie points and dense cloud

[Giada G.]

Hi all,

I am quite new to the photogrammetry but I need to afford a big challenge: reconstruct insectivorous bat skulls in 3D.

Here my equipment and camera settings:

Camera: NIKON D5300 (24 megapixel)
Lens: NIKON macro 105 mm
Settings:
4-5 cm from the specimen
                100 ISO
                f/32
                0.62 sec exposure time
                No flash. 3 white lights
                Focus: automatic focus on a manual selected square


The methodology works quite well with bigger specimens (19 mm x 12 mm x 7 mm) and even the teeth on the upper jaw and on mandible are reconstruct with a decent result.
But with smaller ones (e.g. 13 x 8 x 5 mm) the software has different problems in the 3D reconstruction: it is not able to reconstruct the tie points (providing a long scratch of points) or, if the first step had worked successfully, it isn’t able to reconstruct the dense cloud (doing "inside out" or doubling some features such as teeth and zygomatic arch).

The pictures are in focus and the specimens in the picture results big enough I would say. But please see the link with the examples.
https://drive.google.com/drive/folders/0B0Ja2LUOrW4TdVprX2ZEWTUwREE?usp=sharing

Anyone of you have already tried to perform photogrammetry on such small objects?
Do you have any suggestions, guesses, comments or ideas why the reconstruction doesn't work with this setting?

Many thanks in advance. I appreciate any feedbacks.

Giada

[Kiesel]

Hello Giada,

I have not done this by myself but others. For example Mark Florquin scanned insects very well, see here:
http://www.markflorquin.be/store/nature/6-insects-vol-2/

You find a tutorial for makro scanning by Heinrich Mallison here:

https://dinosaurpalaeo.wordpress.com/2014/09/30/a-more-detailed-take-on-pinus-macro-photogrammetry/

From your images I think that your object needs to fill your image a little bit more and you can put something structured under your object to help PhotoScan with the alignment.

Hope it helps.

Karsten

[mrc]

Hi Giada

I am doing exactly the same thing as you and experiencing exactly the same problem. When I do get the bat skulls to align, they often turn inside out and it is a problem I don't have with skulls of other animals. I suspect it is something to do with the curve of the cranium and the slight translucency of the skulls. Even using some kind of structured backing sheet I get an ok alignment, but the shape of the cranium is completely wrong, it comes out lumpy. I can't get good teeth meshes at all - but I am not surprised by this as teeth seem to be a big issue in general with photoscan.

I have used a few different cameras, and used a microscope camera with focus stacking, but I seem to always get a bad result with bats. It should work, I have recently given up and moved on to other species to revisit the bats later.

Keep in touch and we might be able to come up with a solution together.

Cheers

Michael

[Alexey Pasumansky]

Hello Giada,

Sometimes we receive reports regarding similar issues like you are describing related to the macro photography of the tiny objects.
If you can provide the original image set to support@agisoft.com, we'll also use it in our studies related to this problem.

Also you might be interested in the following article on a similar object scanning:
http://www.hogton.com/focus-stack-reverse-lens-macro-photogrammetry/

[Giada G.]

Thank you all for your feedback!

Michael: we should definitely keep in touch and, finger crossed, we might find the solution. I will try other combinations of chunks and number of pictures. And I let you know if I can reach better results. At the moment I reconstructed successfully some very small skulls (13mm x 8mm x 5 mm) using the 60 mm macro lens but the resolution is quite low (the teeth and the zygomatic arch in particular)

Alexey, I sent to the email address provided two dataset: one made with a 60mm macro lens and one with 105 mm macro lens. I look forward for any feedback and if you need other dataset just let me know.


Best wishes
Giada

[mrc]

Great!

I have lots of trouble with smooth, conical teeth, even on large specimens. Fossil material works well almost all of the time, but recent teeth never quite work properly, I get dips and grooves across the surface that are not there in real life.

I am keen to see if Agisoft can help, I might send off a set of photos of some canine teeth I was playing around with this week and see if they can improve my reconstruction.

[Lapin]

Hey mrc.

You can send your photos by email Formlabs1@gmail.com . You will receive the processed model and file .psz .

Cheers,
Steve

[merckxiaan]

maybe you need to exeriment more with how close you are to the subject. it might have to do with distortion... you could try agisoft lens, altough i've never really used it.

[Skel]

Hey everyone!

I had this same problem, also with bat skulls. What works for me is using the 60mm macro lens, and stacking the photos. I was able to get clear photos and successfully made models. The photography process takes much longer (about an hour for each angle you do), but you get good quality models out of it!