3D for Crazies

[greybeard]

The recent posting which mentioned the toy nailbed, as used to give a 3D metal version of your face(ugh) got me thinking yet again about methods of extracting z axis information.

There are various diy(almost) ideas with links already in the forum, and no doubt they will get posted here also.
While waiting for some of them to come to fruition, I thought I might as well start a thread for other crazies like myself for whom ideas are easier to come by than cash.
This is not to put off contributions by eminent engineers, etc, but just be warned that extra points are awarded for the intellectual distance between one idea and the next.

Starting with the kid's toy, there seem to be two basic problems in getting something useful out of it.
1. Low resolution, unless you can produce a version with piano wire, and 32dpi is ok for you.
2. By extending the whole area upwards, little is gained except a metallic or reflective surface.
This would at least give you a surface with a consistent property to be exploited by the scanning method.

If the toy was reduced to a single line in the manner of the tool used for copying wood moulding sections, this then allows the raising of each contour above the level of the original.
Sideways scanning of this would allow a fairly simple optical method of retrieving the shape of the silhouette against an illuminated background.

A method like this has the possibility, given the resolution problem, of speeding up the production of z axis information, but still requires software to convert the file to a usable form.
Some interpolation might be acceptable, of course, and if you run your machine in raster mode, the programme might be easy to write.(Beyond me, naturally, but I'm sure someone will tell me how.)

My off the wall idea is to paint the object black, place it in a shallow dish, then just cover it with diluted milk, and photo it.
This will give me in a single step a greyscale image of the whole area suitable for converting to a raster carving programme.

Your turn.

John

[CJL5585]

John,
Really like your idea of the milk for acquiring images. Here in the US one can purchase ORGANIC diluted milk as the real thing. It's sold as 1%, 2%, or 5% fat content.

My idea:
I think that if someone could build a STEREO digital camera with electronics that could calculate the difference in the different pixels from the two images and would give an output for the X, Y, and Z coordinates, that we would be in business; That is, if we could afford the camera.

Anyone care to take on the task?

Jerry

[greybeard]

Hi Jerry,
How about moving the camera between shots, or does that introduce too many possibilities for inaccuracies to creep in ?

Or split the view with an optical setup (binoculars plus mirrors/prisms)into a standard digital ?

Organic milk - I drink nothing else unless it's got ethanol in it.

I'll actually try the milk idea tomorrow. For now it's nightcap time.
John

[CJL5585]

John,
With a stereo camera, (it would have 2 sets of lenses and 2 different digital sensors) one would electronically have a true 2 1/2D view of the subject from one side only, not true 3D as stated in previous post.

Just think of the old stereo photographs of WW1 or WW2 days that the military used for intelligence gathering, and the viewer that was used.

Only a single photo would have all information needed to compute the topography for a 2 1/2D dimensioned part. The resolution could be near 1/1600th of an inch if the pixel layout was 1600 X 1600 which equates to 2.6 Megapixels. Many reasonably priced digital cameras attain 8 Megapixels now.

The camera would not have to output any photos, just process the info into a data or pointcloud file usable to create 2 1/2 D g-code.

I Have not given any thought yet to rotating an object and having overlapping photos taken. This action would require custom software which probably already exists in Military and NASA circles, but is probably not available to the general public.

Here is a website relating to a camera setup: http://www.ledametrix.com/
The crossed viewing is by crossing your eyes to see the true 3D images.

Jerry

[ViperTX]

http://www.camtronics-cnc.com/

Before you all re-invent the wheel....take a look at this website.

[CJL5585]

Originally Posted by ViperTX http://www.camtronics-cnc.com/ Before you all re-invent the wheel....take a look at this website.

Oh, I don't want to re-invent the wheel, I just want to go from a wooden wheel to a radial low profile tire with custom rims.

[CJL5585]

"Like doing jigsaw puzzles in the dark"

John,

Could you elaborate on these puzzling things that you are "working on" in the dark?

Jerry

[greybeard]

Jerry, what are you doing up at 6 am or thereabouts ? Doing jigsaws in the dark ? No, I haven't tried it either, but it might be quite a challenge.

Well I tried the milk photo, examples attached. First the object - part of an old drawer handle. Then the photo immersed in 1/50 milk in water. Then a screen shot of the greyscale imported into Designer, the Carvewright software.
I didn't bother to spray it black, not that would have made a vast amount of difference. The crucial oversight, simple afterwords of course, is that the milk obscures detail as well as adding the desired to the greyscale.

I'd earlier thought of using a dye solution, and converting the depth of colour to a greyscale, but the milk seemed easier.
Anyway that's the next experiment, but will have to wait for a couple of days.

[CJL5585]

John,
When you get something that works for you, I want to know the secret since I need the same setup.

My son gave me a NEW FACTORY turnkey servo router system for christmas. He paid $3500.00 per the receipt. It is setup with 3 ea. Gecko 320's, 3 Pittmam servos with Pittman 500 cpi encoders, 8 TPI leadscrews, 2 120 volt switched AC sockets, and Everything just plugs togather, has limit switches, all cables etc. but some weird Supercam control software. I had it running under TCNC awhile ago, but it needs some setup and so forth.

He also gave me a new un-opened dremel tool and a new un-opened wet/dry vac that he had received in Christmas's past.

He is an outstanding human being. I taught him well, and it stuck. I am really blessed and also very thankful.

Jerry

[greybeard]

Originally Posted by CJL5585 John, When you get something that works for you, I want to know the secret since I need the same setup. ....... some weird Supercam control software. I had it running under TCNC awhile ago, but it needs some setup and so forth........ Jerry

Well well. That's the same software that I'm in the process(lengthy as ever) of setting up.
I picked it up in the fall last year, so I suppose a year building isn't too bad by zone standards. The package was as per their kit - software, three steppers, a control box, cables etc., but no hardware as far as the machine goes.
The guy I bought it from threw in the computer that the software was loaded on as well for free. Like you say, it's designed to just be plugged together.
However the software is unusual compared with what most people use on the zone, with the g-code support seeming to be as limited as the designer thought necessary. Mine's the 98 version, and I see no season to lash out for the xp yet. Certainly a raster mode seems to be the way I'll start.

Back to 3d.
I'm next going to try immersing the same object, after I get some matt white paint today, into copper sulphate solution. This is the one simple innocuous solution that I can think of that doesn't dye the object(Even meths has a tendency to leave a purple stain).
I've just thought of using malt vinegar, but that would probably attack the metal.

It also occurred to me that if I put an object in a glass beaker of solution, resting on a turntable, I could then produce a complete 3D panorama of the surface. Having done the same with a plain cylinder as a reference, I could then remove the errors produced by the container.

Now the tricky bit, where we get to the "jigsaws in the dark" - what do I do with the images. I've no idea how to get from there to a 3D mesh/g-code, or whatever is required.
40 years ago I wanted to be a programmer, but never pursued it. There you go.
John

[epineh]

OK, I will bite, what about triangulation, three laser beams (what else???) very small resolution, software to move the beams and recognize when all three are all aligned (I never said this was gonna be simple), calculate the beam's relative angles and use trig functions to calculate depth... simple

Another method is to utilize a standard scanner, install a wing wong mounted to a doovalackey that will interpret the signals sent by the scanner head into a depth value. OK so the second idea isn't alltogether serious, but it is a concept nontheless.

Failing that you could use latex rubber, immerse the part to be copied face down into a pool of the stuff, wait for it to dry and use it as a mold.

OK... I got nothin'

Russell.

[greybeard]

Hi Russell.
Your contribution is in keeping with my original intention of the thread.
Please keep us up to speed on your development of the doovalackey.
We had a wing wong, but it died.

I, too, wondered about using a scanner, but mine doesn't like working upsidedown, so I thought about converting the milk into a jelly, then I could turn it out onto the scanner glass. Not sure about the effect of lowering the lid, though.
Regards
John