3d scan help? …

I would like to have some objects 3d scanned. I actually only need the 2d profiles. I’ve done the scans on my copy machine but they never come out right. Does anyone in the group do this or know of a good place to have it done?

Flatbed scanner only works well if the entire profile is laying on the glass. The camera is in the center of the carriage, so anything away from the glass suffers from parallax error.

I recently did job that needed to be scanned. I found several local businesses that did it. It’s not cheap.
Luckily a friend of a friend, and former customer had one in his garage & did it for me.

My son told me there are apps that can do it from your phone camera. No idea how it works or how accurate it is.

Google maps is your friend!!! :wink: Just try a search for 3D scanning.

Scans on the copy machine should come out fine — see the technique at:

I’ve scanned profiles before, and they are usually very close but when you are making knife handles they have to be perfect. I’m just searching for a better way.

Use a combination of scanning & reverse engineering (measure & manipulate using scale, offset, etc.)
The scan will give you the shape, then you scale it to get the dimensions correct.

I’m guessing you are placing the flat side of the knife scale on the scanner to get the outside shape, and possibly any pockets or holes. You can improve the original scan by painting or coloring the object to increase the contrast. For example, paint the bottom side of the object white, then paint or ink the face that contacts the scanner black.

I’ve also done this with my phone camera. You need to make sure the object is sitting as flat as possible, and the phone camera is directly above the center of the object, the greater the distance, the less parallax error. (the edges of the object are just a bit further from the camera than the center)
Then you measure the object to align & scale it once it’s traced.

1 Like

The thing is, a scan should be very accurate — scanners have a very high optical/mechanical resolution and interpolation increases that quite a bit — it’s just a matter of getting a good quality scan.

One other thought, cut your scanned image on a piece of scrap first. Then if needed you can always node edit that scanned image for any slight adjustment and test it again.

1 Like

Another consideration is if using a belt-drive machine, make sure that the X- and Y-axis are calibrated:

This is usually the approach I take, it’s just very tedious.

1 Like

And to throw another idea out there, If you’re looking to do this yourself on an ongoing basis the openscan project looks awesome.

I’ve been following it for awhile and the resultant scans I’ve seen are incredibly accurate. Given the price of entry, it’s my top recommendation for folks wanting to get into 3D scanning.
Link: https://en.openscan.eu/
(Their subreddit is also great)

Though I’ll echo what Will has said, if you are only interested in a 2d profile, a flatbed scanner should be way more than adequate and considerably easier.

I’ve read some articles regarding touch probes used for this purpose. Is there anyway one could take this approach?

Touch probe could work and there is a good amount of prior art on the forum to reference.

The issue you’ll run into is in the fidelity of your result. You could imagine each touch off of the probe as a unique point in a final point cloud.
The more points you have, the more accurate the result.

For reference, this is ~how face unlock works on modern smart phones. They use a sensor to generate a point cloud of your face with many thousands of points…which is great and super accurate but it would take an eternity to generate the same number of points with a touch probe.
(You’re probably better off finding an app that lets you generate and export point cloud data from a phone than trying to leverage a touch probe…At least for anything more than simple contours)

See:

Honestly that’s over my head, I was looking for something I could buy. At my dads machine shop, he has a probe that goes into a mill and will basically trace the part. The software is compatible, is the only problem.

When I needed a high precision tracing, I printed a fine grid on a piece of transparency film, put it, and a calibrated square on the scanner bed, then put the part on top of the transparency film and scanned that — the additional positional information of the grid squares made it possible to get things lined up just about perfectly.

The other thing to do would be to make a device for tracing around the part — drill a hole in a piece of stock which matches the diameter of a pen refill, then cut an acute angle up to it on both sides to cut it out so that the indentation of the hole is at the tip of an acute triangle. Tape a pen refill into that indentation, put the part down on a sheet of grid paper and trace it out — use a sharp pencil to trace where it is convenient so that you can see the offset of the radius of the pen refill. Draw along the center of the pen line, then inset to where the pencil line shows the actual edge of the part.

2 Likes

This topic was automatically closed after 30 days. New replies are no longer allowed.