Probe Routines

[davidp]

Hello All,

I have been commissioned to produce panels for a yacht refit. For all of the panels the shipwright has produced a template and I have to copy the template and then cut the corresponding panel. They are outline tracing scans not 3D probes.

Using SB Copy Machine to probe the templates is fine but as there are a lot of straight edges on the templates and some of the templates a very large and often it is just the transitions and certain sections of the template that needs a high resolution scan with the straight sections just joining the dots.

Is it possible to adjust the resolution of the scan during the scan? I have not, as yet, found a way to do this.

If this is not possible, is it feasible then to scan small sections of the template at a high resolution and then join the scans? If so how does one get the probe to move in the right direction at the start. It always starts to move in the Y + direction, to scan different sections, I will need the probe to begin its moves in the Y+, the Y-, the X+ and the X- directions at various times.

Any and all advice would be greatly appreciated as I will have over 1,000 templates to scan.

Regards,

[Gary Campbell]

David...
When doing a similar project, Brady Watson explained a tecnique where you take a digital picture of the template (or piece). Also in the picture would be an item of known exact size, say 12" square or circle for cabinetry sized templates. Convert the .jpeg to DXF. Select the known object and scale it to match its known size. Note the scale percentage. Cancel and scale entire DXF by the noted percentage. You now have full size vectors.

To keep symmetry and size keep camera centered in view field and take only straight on shots.

We did this with actual full size panels and had to play with backround colors. As usual, Brady was right on the money with his suggestion of a dove gray. If this works for you, thank him as I did.
Gary

[nailzscott]

I hope you have better luck than I did of taking a picture and keeping the symmetry correct. I took multiple, straight on pictures of a 2' x 5' shape and tried to match a potential customers existing template. No amount of sizing and scaling could match the customers object within an 1/8" on the perimeter edges. Admittedly it was an oddly shaped object but after 3 hours, I just cut out what I had. Fortunately in this case, he liked mine better than his original template. It may depend upon how precise yours need to be. Good luck.

[Gary Campbell]

Scott...
We went thru the same things. Luckily I was advised to use the gray backround and set non glare painters lights on all 4 sides and shot pics without flash. That is what it took to eliminate shadows caused by the flash that made the part offsize. Hopefully David will be able to take a pic of a template much easier as is doesnt have the shadows associated with plywood parts.
Gary

[Brady Watson]

Getting accurate scans via photogrammetry isn't too difficult, but you must take care when doing it. As Gary mentions, a grey background placed behind the object to photograph makes a big difference. You also want to AVOID USING A FLASH and have good diffused lighting while taking pictures. While it's tough, but not impossible to hold the camera steady & true in your hand, you'll want to pick yourself up a cheap tripod that you can use to setup the camera squarely to the workpiece. I've seen these for as little as $15 at Wallymart.

After you have acquired your 'scan', bring it into (I'm guessing PartWorks2D - or ArtCAM) and ROUGHLY trace out the outside shape of the part using ONLY lines until you have gone around the complete perimeter. Then in node editing mode, convert those straight spans to bezier/French curves and fit them to the outside of the shape, adding nodes where necessary. Avoid using arcs.

Depending on how crisp the picture is, you can use the autotrace bitmap tool in PWorks. From a professional point of view, I rarely use this since it often results in parts with many jagged nodes. I prefer to do it by hand & the resulting vectors are completely smooth and ready to machine. Your software skills really make a big difference when doing photogrammatical tasks such as this.

-B

[ljdm]

Brady - what is the reason to convert to bezier curves instead of arcs? I've done it both ways, wondering why bezier is better? Thanks.......

[dana_swift]

Lou, a bezier curve can make an arc, but it can also represent more than a simple arc, so it makes a better choice as Brady has noted. Bezier curves are as much of an improvement on curve generation over arcs as CNC is an improvement over manual building.

I have seen explanations of bezier curves for artists, which stay away from the mathematics and mechanics so mere humans can understand them. Once you see how to control the curve generated by the four points it is one of the most powerful and easy to understand graphical tools ever invented.

A bezier can create a straight line. Interestingly a bezier cannot be a part of a true circle, but it comes so close humans cant tell the difference. Two bezier curves make a figure that looks so much like a circle that unless you are from the math police its safe to use it as one.

A single bezier curve can loop around into bowties, make S figures of all manner of curvature, and make V shapes with graceful rounding at the peak. A sharp V is not usually possible with beziers tho.

PW has bezier manipulation tools in the node editing mode, once you get the hang of how to control the curve shape you will be able to generate any arbitrary shape, arcs cant do that.

And when you have mastered bezier curves, then you can consider working with NURBS which PW does not support, however other software does. NURBS are a step above beziers just as beziers are a step above arcs, and real arcs are a step above straight line approximations to an arc.

(NURBS stands for "non uniform rational B splines" if that helps any!)

Hope that was the right amount of info-

D

[davidp]

Hi Guys,

Thank you all for your input. I have tried photography and it did not work well enough or quickly enough for what I need.

I have adjusted the probing routines to start in a nominated direction ie x +- or y +- instead of the y+ originally.

I now probe the entire profile at something like 30 mm res and then go back and probe the "delicate" areas at a much finer resolution and then join the scans. So far while not a perfect solution it is working ok

Once again thank you for your ideas and solutions

Regards,

David

[Brady Watson]

David,
The probing routine will initially be faster than 'tracing' until your drawing skills get up to par. Practice makes perfect...

-B

[davidp]

Hi Brady,

Even with practice nothing is fast enough :-)

The quickest way is to trace the templates in AI using its trace function, manually adjusting that trace, using both AI and Artcam, then toolpath. However as Scott mentioned this process is not accurate enough, one, because the templates are so big and two, the photographs are distorting too much to produce accurate results. (we only have a tolerance of +- 1.5mm ). We are using a Nikon D70, studio lights and tripods so the equipment is not an issue, the quality of the output is the issue.

I may not have made myself clear enough at the begining but I was hoping that some one would have written a utility to start the probe in different directions instead of just the Y+ and to specify a section of a template to probe instead of the whole template. This way I could vary the resolution depending upon the detail of the template and just join the dots for the straight sections.

I forgot to mention that the material we are cutting for the cabinetry is a new composite material and is approximately $800 per sheet, so any mistakes are very costly.

Unfortunately it appears that I will have to refine my own routine to make it work better.

Thank you all for your input.

Regards,

[bcammack]

David,

See if there are any big, automated countertop fabricators near you. They might have a 2D, flat-panel digitizer (as we do) to convert the physical templates into accurate CAD representations.

Ours came from these people: http://www.outlinetechnologies.com/index.html

Perhaps there's a similar unit near you now.

[davidp]

Brett,

Thank you for the suggestion. I will try to find some one with a digitizer in our area. Shame the freight is so expensive other wise I could give you a job:-)

We have completed about twenty some templates and cut them so far, the largest error so far has been less than 1 mm so fingers crossed it is working.

Only 980 to go?

March can't come quickly enough!!

Thank you all for your input, while you may not think so you have all in some or other way helped.

Regards,

David