Wednesday, November 25, 2009


3D Scanner


I saw an awesome article on wired. I was wondering if anyone has considered mounting a webcam next to the print head and using it as a 3D scanner. Maybe enabling adding features to premade objects. Scan in the object, manipulate it in our design software, add a feature and then print this feature onto the object.

I've been thinking about different scanning methods. The webcam method relies on the object being scanned having a fairly flat surface texture. I do not think it would work well at scanning parts printed by the reprap. Other objects might have similar problems.

It might work better to mount a laser line generator to a moving access and use a fixed camera to triangulate 3d data.

The best idea I have heard makes a lot of use of the exsisting hardware required for a reprap. A pin probe is placed on a pressure sensitive sensor. The probe is moved around in xy and moved down in z until the sensor detects an object.
Maybe a little like this one ??
This is a sophisticated photogrammetry program, an automated version of Photomodeler, for example. I have a program called Geometra that works in a similar fashion, but not automated. I set a project, take a bunch of photos and correlate the features manually in each photo, the program triangulates the distances between them. The arc3D web service produces models from photos the same way. None of these scanners or photogrammetry programs give good enough results that the point references produce smooth surfaces, or don't need post processing in CAD, and I can see this one being most useful for producing CGI keyframing. They're most useful for establishing proportional relationships between features. Apart from Zsurf, which produces .igs surfaces from height maps extrapolated from greyscale bitmaps, (so can do useful things when bitmaps are thoughtfully edited), all of the software solutions that produce curved surfaces from point clouds are either expensive standalones or plugins to Rhino or other CAD apps.
mccoyn: The laser line approach is entirely doable. I saw it demonstrated at HAR2009. One way to do it could be with a fixed camera, fixed object, and handheld (RepRap-mounted?) laser line: and another way is with a fixed camera, fixed laser line and rotating object: TriAngles at .

Both of those are built in a way that doesn't play along with Wine, though, which is a bit of a bummer.
I saw that laser line approach at HAR2009 and was quite impressed by the simplicity of it all. But later I read that you don't need the laser at all, a sharp shadow will do the trick just as well!.
I like the Caltech shadow scanner, that's easy & source code is available.
Things poorly imaged from one angle (as can be seen in their example) have holes. multiple scans could fill this area. Is there any way to merge multiple scans of this (or should I write the software for that)?
Reinout: According to the Intricad guy I talked to about their stuff, the shadow digitizing process is patented, while the laser line process, astonishingly enough, isn't. There are many weird patents around...
Post a Comment

Links to this post:

Create a Link

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]