Best method for capturing all sides of an object?

(Abby Crawford) #1

Hi there. On a project that I'm working on we are performing photogrammetry on excavated artifacts of various sorts (ceramic, bone, metal, etc.) with an aim of both documentation and online publication. As such it is very important that we capture all sides of each object. We use Agisoft PhotoScan and in the past we have created two chunks (top photos and bottom photos, essentially) and merged them. This process is relatively time consuming and does not have a 100% success rate. A colleague of mine has had better success with putting objects on glass risers and capturing the bottom that way. I was wondering if anyone has any other clever suggestions for how to capture objects in the round in just one go.

(Tom Goskar) #2

I use the "two chunks" (or more) approach, and often use the turntable method (white turntable, white background) to capture the object. I have tried glass, but find that the reflections cause more trouble than it's worth.

I have also used "museum gel" - a crystal clear gel used to keep objects still in their cases - but if the lens and photo quality (assuming you're using photogrammetry) are good then it won't work very well. Back to the multiple scan and subsequent registration approach.

Don't talk to me about registering two sides of a very thin flint object together - I spent days without success (depth of field problems to get enough of the other side of the object to register the two sides together)!

(Abby Crawford) #3

@tomgoskar, I can only imagine how challenging a flint would be!

The other day I bought a piece of glass and started experimenting, but as yet I'm not thrilled with the results.

During a paper at a recent conference Effie Athanassopoulos showed a picture of her laser scanning setup for sherds. If I remember correctly it looked a bit like each sherd was held lengthwise between two fairly thin white rods. I wonder if this sort of suspended setup would work for photogrammetry. I'm not sure if it would speed anything up, since the contact points where the rods meet the artifact would have to be shot (and perhaps chunked) separately, but it might permit a less side-to-side approach.

(Tom Goskar) #4

You could experiment with 'invisible' methods, depending upon the fabric of the artefact to be scanned. For example, a small pot sherd could be held in place by a few needles. Now that's an idea. Obviously, conservators would have to be consulted, and safety considered.

Hmm. I'm now thinking of a turntable, a block of polystyrene, and some long sewing needles. Could be worth a shot.

(Micropasts) #5

Hi @tomgoskar and @abbyec, we had similar issues with modelling flat objects (like spearheads), although most of the time we managed to stitch together the two chunks properly. Rob Kaleta, working with us, had a suggestion to take photos of objects on their thin edge. You can view his post and screenshots here: Note the Blu-Tack! smile

(Abby Crawford) #6

@micropasts, that's a helpful thread. Thanks!

@tomgoskar, I'm going to try to craft an artifact vice-type object this weekend and will report back with my results. I just have to find the right tools....

(Sofa Arte) #7

I'm rather new to photoscan so this is more of a question than a solution for your problem, but i'm wondering if is feasible to work with masks instead of chunks. Sure it's more time consuming but maybe you can make the model in one go without the need of alligning chunks.

(Ben Kreunen) #8

I typically use 2 or more chunks for objects. For fragile objects we just sit them on a linen cushion. You can mask the object in the images, but the cushion can help with the alignment step. Having plenty of overlap between the chunks helps to speed up the mask creation process as you can use very rough settings for the masking without having to be too careful at the edges of the object.

For flat objects I'll use something like plasticine or blu tack to create adjacent objects that will provide a link between images along the narrow edge.I use this approach with laser scanning too, which can have similar problems.

(Abby Crawford) #9

@uomdigitisation, is the linen cushion firm (thereby holding objects up a bit), or do the objects sink down into the cushion?

(Ben Kreunen) #10

They're essentially mini bean bags. Supplied to us by our conservation people (so they're happy) and nicely filled for the right balance of softness and support. The object may get partly "buried" in the bag, but then you just turn it over. Turtle in the pic isn't exactly fragile, it was just a test. In this example I shot 4 chunks, top/bottom and left/right. Initially processed without masking, trimming the point cloud in PS and aligning/merging in Meshlab. But with masks you could theoretically run it through as a single chunk.

(Ben Kreunen) #11

Yes you can in theory. The align images step is the one that allows you to ignore masks (can be useful in some situations), but the dense clouds/mesh/texture stages always use the masks. If you have reasonably flat lighting and generate the masks then it may be practical to use that approach.

(Sofa Arte) #12

That's interesting, thanks. I'll do some testing.

(Abby Crawford) #13

@uomdigitisation, thanks for the visual! I'll have to find myself a little beanbag.

@tomgoskar, my attempt at an object holding device was a bit delayed because of visitors in town, but on Sunday I finally had a chance to put something together. There are several things that I would now change about it because it is not exactly right, but here is a photo:

It's just a few pieces of wood held together with corner brackets, and then a piece of machined steel threaded through a hole at the top. The steel piece is much too long and would not remain perfectly vertical as a result.This meant that it would not be suited to holding anything fragile. I fitted a small, square metal piece to the bottom plank of wood where the object touches it because I didn't want anything to sink into the wood. If I did it over, I would likely change the metal piece to a short rod or small cylindrical platform. I'm not over the moon about this first attempt, but it provides a good basis for any future attempts that I might make.

(Tom Goskar) #14

That looks great - a little touching as possible and much safer than the 'bed of needles' approach that I had envisaged!

(Abby Crawford) #15

I spent this afternoon going back and forth to various hardware stores, and have made something that is a bit more user-friendly than the last attempt. It also required significantly fewer tools and effort to put together. It's just 3 pieces of 1/2-inch copper pipe, a copper corner piece and a copper T-piece. The T-piece is held in place by a couple of stainless steel hose clamps that can be adjusted up and down to compensate for differences in object/artifact size. The base is just an $8 piece of pre-cut round wood into which I drilled a hole with a spade bit.

The benefit of the copper is that it is slightly malleable and I can tweak the roundness/flatness of the horizontal ends so that they can hold objects better. The one thing that I'm lamenting right now is that the T piece is a bit too big and tends to wobbliness if not properly pinioned between the hose clamps. Luckily it's very easy to slide the T off if I end up finding one that fits better. I haven't yet tried to shoot in an object yet, but will update when I do. Tomorrow, perhaps.

Of course, now that this one is done I can think of several ways that I would improve it...stay tuned smile

Oh, and @tomgoskar, your bed-of-needles idea is nothing compared to the hardware store employee's suggestion that I just put a hook in the ceiling and get a piece of string or wire to dangle the object. What could possibly go wrong? wink

(Tom Goskar) #16

Very inventive! Looking forward to hearing how well it works.

(3DGEOM) #17

This is super clever! So do you not need full 360 shots around the shoe to get photogrammetric reconstruction? @abbyec

(Abby Crawford) #25

@chrisleggett, You still have to take photos of all sides of the object. As it turned out, the copper contraption that I made was not very useful at all. It was worth a shot, though!

(Gerpho) #26

I use Acute3D, so that I don't know if this method would work with photoscan.

I place the object on a white turntable, and shoot a series of pictures (30 at least) with a full 360° rotation of the turntable.
Then I put the object on its side and make another series of 30 photos over a full 360°. The only tricky thing is to make sure that there is an (almost) common view between the two series.
Here are the two starting positions of the two series of photos for this little bronze sculpture :

If this is done properly and if the "common" view has enough features, the two series will connect automatically, and the result is here (notice that the engraved base is visible) :

Tiny bronze sculpture by Béatrice Bauchart by Gerpho on Sketchfab

(Panomedia) #27

Hi Everyone I use BobCapture with looks much easier than all the hassle of Agisoft with the chunks, unless i haven't understood order to get the bottom/base of an object : done in (BobCapture). i was just wondering if it can also be done with agisoft ?? although your bronze sculpture on Acute3D does the job too smile ...

How to merge the base/bottom of an 3D object on Agisoft?