Official Luthiers Forum!

Just wondering, mainly for discussion purposes, whether there is any interest in scanning existing instruments as a means of "Reverse Engineering"?
http://www.david-laserscanner.com/
These guys make it sound rather simple with a hand held laser "swiped" across the object while a webcam captures the scan lines into the software.
Not sure whether the scan would have enough accuracy, though.
Nelson

Well,

like anything in life, you get what you pay for...

I actually build this a few months ago, and for "small" objects the results were "ok" but nothing to write home about. I know their site makes it look very easy, and with a good quality green laser, you can get "somewhat" acceptable results....but it is not a great package/solution IMHO for larger objects, including guitars.

The other thing you need to consider, is that with the shinny surface of a guitar, the laser will bounce right of it, and create reflections. These will show up as spikes in the model, which will later require some heavy post-processing work with a reverse engineering package, like Geomagic studio for example.

Folks create solutions around this, by wrapping the guitar in plastic, then using some sort of paint over it. I have even heard that some folks use "foot powder" or some other type of spray, to get the shine off, and then perform the scan. Again, all results I have seen with it for large objects, do not look that great. I guess your mileage will vary.....

There is a reason why higher end scanning solutions are in the 30-50K range....

_________________
-CyborgCNC
http://www.cncguitar.com
https://www.facebook.com/CncGuitar

I saw somewhere that a University was making an open source scanner/software package out of an X-Box Kinect. I believe the project was supposed to be easier and the results better than the laser scanner. I can't find the original link, but google Kineckt 3D scanner for a bunch of different links.

Bob

I've used commercial laser scanners for reverse engineering instruments, and I've also played with the David software a bit. I had a high end laser from a consulting job where I used it for calibration, and an HD webcam, and the results were quite usable with everything dialed in.

Laser scanning is all about calibration, and that's what you're paying for with expensive scanning systems. The David software is capable of doing high end work, but you will need to invest in quality hardware and be extremely picky about your equipment and setup. The algorithms for stationary laser scanning are old as dirt, so your results will come down to buying top notch components and dialing in your setup and procedures. Having really consistent laser motion helps a lot as well, so make some sort of linear or rotary system to move the laser if you can.

I think of laser scanning as being an intermediate step on anything non-ornamental. Even with Geomagic or Rapidform a good result takes more time in software than modeling from scratch with the scan data as a reference.

The issues with reflective workpieces are spot on, so you have to decide on a way around that. Dusting is the least risky but it tends to have 'spiky' data with high and low spots where the original surface shines through. Masking tape works pretty well if you're really careful to lay down a flat and uniform layer. I've got a feeling there's a way to account for it with optics, but that's firmly out of the realm of DIY solutions.

The Kinect stuff I've seen is much faster than laser scanning, but it's all been much lower resolution. Really cool for real time scanning and quick scans, but so far as I know nobody's modified it to get precision scans and the data might be too noisy for that to come to fruition.

_________________
Bob Garrish
Former Canonized Purveyor of Fine CNC Luthier Services

I tried the David software as well. as others have said, very mixed results. Red lasers, don't waste your time with it. A good green line laser and the results are much better. Most important thing I found was the need to control the feedrate of the laser. This sounds easy since we have CNC, but with the calibration boards, the need for darkness, and other things in the equation makes for a real PITA. Just not worth it to me. I can get a much better point cloud using a probe.
MK

_________________
I'm outside looking in, just farther from the window than most.

Thought I'd show an actual file..... laser scanned......

This is an actual neck scan....un-touched..... converted into an stl file then brought into Rhino and converted to a jpg.

kevin

And there Bob is why I ask you this question! I have had three of my products digitized via laser, and have been underwhelmed by the results, and this is by two different systems....one a 40k version and the other one of the less expensive table top versions set to really low resolution. Honestly the results were not that much different and I have ended up remodeling them from the point on the scan. Well, actually, myself and two others have modeled them.

Now my question. Would it be possible to use a laser to obtain specific points like you would with a Faro arm so you could use those point to make the model. Laser scanning produces so many points its crazy, the model is heavy and I hate stls! I ask this in sincerety knowing your backgound of being a math guru! (edited to add) I forgot to mention maybe using a small web cam for positioning the cnc as you obtain points with the laser.

Mike

I've done a fair amount of CNC probe scanning of my carved plates but it's a slow process at best.
Could be much faster if there was a better digitizing plugin available for Mach3.

HI Nelson,

Actually there are two digitizer scripts for Mach3 (older version) over on the Mach web site. You will have to search for them, but they are there. One was done by a gentleman in England, and worked well, and the other was done by a gentleman in Florida. Both worked and worked well, but both a little different. I can't remember which one, but one would trace the outside profiile of an object, then only digitize within that boundry. Sped things up considerably.

Mike

Hey Mike,
I've run both of the digitizing scripts and the first (old) one works good but really slow because the Z has to retract to a safe level after every hit. The second script you mention that first digitizes the outside of the object showed a lot of promise. I could not get it to run on my PC/Mach3 and talked back and forth with the author several times who was very willing to help. I'm running a different PC now so maybe ought to give it another try. Sounds like you were able to run it successfully.
Nelson

http://www.flatbed3d.com/
http://www.machsupport.com/forum/index.php/board,67.0.html

Here's a little info on scanning that may (or may not) be of interest.
Guess it would make sense to use a CNC machine to position the laser incrementally.
Sounds like the camera could be mounted to the same "carriage" since it's using triangulation to measure height differences.

Nelson I ran, and still run Mach2. I used the original Art F designed digitizer in Mach 2, but a preset default caused me to not be able to actually use the other scrips for real world point taking. The Mach 2 default original positioning move, drove my digitizer probe down into the table, bending the shaft, and slightly warping the little plactic shaft holder. I did test both Mach programs, with some success, but because I could not trust the accurcy of the digitizer itself, I have never used points obtained from either script. I have them saved on my pc somewhere, or at least a pc, and will try to pull them up to re test.

I would really like the web cam laser system for obtaining individual points. Manually grab all your outside primeter points and then let it do it's thing automajically everywhere within those points!!!

Mike

Mike,
Let me know soon as you get that up and running and I'll be there.
What kind of beer do you prefer?
Nelson

Nelson, I drink ginger ale! Vernors to be exact! That said, the sad thing is I know a guy that could build this laser/webcam thing in his sleep. I don't know if you remember over on the cnczone a few yrs back a guy introduced a cncbrain! Well, I am one of the purchasers who have yet to be able to use it due to lack of correct software, but I have kept in touch and have been very supportative of the guy because, deep down, I know where he has been, and know that somewhere on the horizon, there awaits something really good from this guy.

He and his team used a quasi AI and video recognition software, which he wrote, and won the military's two unmanned vehicle competitions out in the desert! Something as simple as we want would be like child's play, but I can't convince him to waver from the path he is on long enough to do it.

I also belive our very on Bob Garrish could probably do it.......eh Bob....wanna try?

Mike

Sorry for the late response, Mike, I actually had one mostly written up and got so busy that it sat there for two days and then I accidentally closed the browser! It's a busy week, trying to get my masters thesis defended before holidays.

Getting individual points is as easy as culling the cloud from a laser scan. As I understand it, the entire line is required to map the points on it into 3D space. Getting 3D data from a single point would require some crazy hi-res optics (dot size) or dual cameras. The reason normal laser scanning needs only one is that the laser and the camera have an angular difference, so the algorithm is working on the presumption that one 'camera' (the laser) always sees a straight line.

The FlatBed 3D system looks to be using a very simple algorithm. I don't think it would be too hard for a hobbyist to set up / code a similar system with very basic image recognition software. That won't likely be me, though, as I'm going to have a lot on my plate for the next few years. I need to build/integrate some sensor systems and code AI for an autonomous robotic submarine.

With the right optics, and a limited working depth, a 'dot size' system wouldn't actually be too hard to realize.

_________________
Bob Garrish
Former Canonized Purveyor of Fine CNC Luthier Services