In regards to a depth sensor, there's a good change Aram is referring to 
a Kinect or the Asus Xtion (Pro Live). There's been plenty of 
hacker-friendly attempts for "point-n-shoot" capture solutions using 
these over the years, adding to the laser-line, and distributed-light 
methods. The Xtion unit retails sub-$250, so it's an expensive 
experiment, but a contender for entry-level scanning - but it's only 
part of the hardware side. The DAVID project uses just laser and 
distributed light methods (at current, IIRC), so not a compatible 
software piece. There are retail packages that speak with a Kinect or 
Xtion, my favorite low-cost at the moment is Manctl's "Skanect", but I 
also like the free Faro (tickler), "Scenect". There are of course 
others. Most are Win32 or Win64 offerings only. OpenCV is often used as 
a framework for open-source attempts, so not all hope is lost if someone 
wants to try....

In a commercial environment, I do create and provide "difference scans" 
of parts in some cases as required by a client, but it's pretty rare. A 
big thing to note about volumetric scanning versus probing is that 
scanning gets you a profile, and it takes a lot of time; probing gets 
you parametric data and can be extremely fast.

To implement an in-process scanning technique effectively, I would have 
to do it in between machining operations - not just at the end, as one 
part feature may be dependent upon the next, such as a helical thread 
inside a drilled then bored hole) - and although most faults like that 
would cause tool damage, I wouldn't want to write part programs so that 
each individual feature is a separate operation (that's what automation 
is supposed to make easier!) - thus the logic decision as to how to 
"fix" the problem won't exist.

Even if it did, one would have to scan a "known-good-master" at each 
step of the production for the test scan to reference against. If you 
want accuracy, measuring scale on a single volumetric scan is near 
impossible. You need to have many angles and views stitched together to 
see past obstructions. Any imperfection in the part, including 
unintended, is recorded [perfectly].

Furthermore, the scanner is unable to understand chips, swarf or coolant 
sitting on the part - it will look like a fault in the part and raise an 
alarm. Get coolant on the lens and your scanning is of no value.

Probing, on the other hand, can be more easily automated, and directed 
towards achieving the stated goal - whether or not a particular feature 
is within tolerance.

I use both laser and depth-based scanning plus probing for a variety of 
tasks, but only probing on my machines.

I would liken volumetric scanning versus probing to the earlier 
implementation of a webcam (camview by pavel et al, for example, which I 
really do applaud) for tool-length-setting (and more) versus a touch 
probe. Although I'm enamored with having a non-contact tool setter, a 
camera is much less tolerant of mistakes or changes in the environment 
than a touch probe. When you just want to get parts out, the touch probe 
"JustWorks".

That's not to say that having a volumetric scanner as a tool in a 
machine isn't a possibility or a value, but the historical intent for 
LCNC not to require a "64bit 8-core watercooled machine with Cray 
loadbalancing (sic)" makes it a difficult challenge to have it 
integrated with LCNC. My preference would be to have it on a separate 
machine as a standalone application, but load the sensor as a tool when 
needed. The benefit is the automation of the motion of the camera 
(instead of human hand) which in that aspect, would return a much more 
accurate result. Since the software has to "chew" on the scanning 
result, it would actually be dangerous to have this type of interruption 
(and I guarantee it would be an interruption) to the safety of the LCNC 
realtime loop.

Ted.

------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:

Build for Windows Store.

http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to