3D Scanning versus 3D digitizing

I have been following the 3D scanning with some interest and have had a look at the David scanner. There seems to be some discussion as to the resolution of such devices. there is also a fundamental issue with the output of such devices.

One of my other hobbies is animation and most 3D scanned models are of very little use for the following reasons.

The vertex count is often too high.

The vertices are generally positioned incorrectly on the surfaces. This is often vital if the model it to deform for animation or if you want to vertex map an image over it. "UV mapping" is done to texture the model and to ad colour. A scanned model requires so much post processing That you might have well modeled it from scratch. Such models often sit in a repository and no one actually uses them. Even static models benefit from this. For example a model of a car is only useful if the vertices are positioned around the edges of the windscreen, wheel arches etc. Otherwise you can't fit the correct materials to them.

As a general rule the best way to get a real life object to a useful 3D model is to digitize the 3D points. That way you get exactly the points you want. It takes a little longer at first, but pays dividends later on when you actualy want to do something useful with the model.

Has anyone considered making one of these as illustrated below. I think it would be easy to reprap and using a few encoders would be a nice little project. I might have a go myself so if anyone can suggest any decent encoders or experience with mouse encoders, perhaps we can start a discussion here.



Edited 1 time(s). Last edit at 12/21/2010 08:25AM by martinprice2004.

One of my other hobbies is animation and most 3D scanned models are of very little use for the following reasons.
Never thought about that; I'm more keen on the sculpt->scan->print side of things, but for static figurines. For that I think I can get by with camera and laser-line scanning.

Still, sounds utterly cool!

I hereby dub the project Mister_Probey.
(code name / working name)

What sort of positioning accuracy do we need for the probe?

If we mount a 200 mm rod on a rotation sensor it will have a positioning resolution of

resolution = 200 mm * sin ( 1 tick mark)

so ....

>>> 200 * math.sin (2 * math.pi / 4096.0)
0.30679603725695309

for 4096 points / rotation, about ~0.3 mm

---

>>> 200 * math.sin (2 * math.pi / 2048.0)
0.61359135259319519

for 2048 points / rotation, about ~0.6 mm

---

>>> 200 * math.sin (2 * math.pi / 1024.0)
1.227176929830895

for 1024 points / rotation, about ~1.2 mm

---

I'd suggest a

---

Has anyone considered making one of these as illustrated below.
Yup.

I think it would be easy to reprap and using a few encoders would be a nice little project. I might have a go myself so if anyone can suggest any decent encoders or experience with mouse encoders, perhaps we can start a discussion here.

Take a look at http://usdigital.com/products/encoders/incremental/

or possibly one of these and some printed herring bone gears to gear up.
http://techartblog.blogspot.com/2009/02/cheap-absolute-optical-encoder.html

---

-Sebastien, RepRap.org library gnome.

---

Remember, you're all RepRap developers (once you've joined the super-secret developer mailing list), and the wiki, RepRap.org, [reprap.org] is for everyone and everything!

Check out: [www.3ddigitizing.net]

How does the point sampler determine where itself is?

I think the best way would be to combine a point sampler with a line generator laser. Could this be done by measuring the reflection time on the laser return... I guess that would need to be so accurate that wouldn't work...

We use something like that where I work, except ours is 12ft long lol. The one we use is by a company called Faro. I'm not sure if that helps at all.

Jago, the way my boss explains it, is that each joint measures the distance traveled and uses the values from each to determine where the tip is. You can take it from me, when they're calibrated, they are extremely accurate!

Here's a clip showing my boss (dude with the headset) and my old co-worker scanning a Bugatti Veyron (2:18). [www.youtube.com]

So, from my personal experience having worked with laser scanned data, CAD data and data from a scanner like the one above, I don't know if this would be the best for printing. This sort of thing is great for visualization (video game and print models) because you get a clean and efficient mesh but its not really ideal for other purposes. A laser scanner can capture every detail and can later be brought into a program like TopoGun ($100) and re-surfaced and optimized if needed.

Whereas a scanner like this, you have to draw the mesh on the object at the resolution you want and scan it in by hand. If you find out later you missed something, you have to set everything up again, recapture the missing stuff and align the new data to the old data.

Don't get me wrong, I'd love to have one of these at home to play with, but I wouldn't likely use it for printing stuff. Ok, well I might, but I'm a 3D modeler by trade (6 years) and I can add in any missing detail in software lol