Machine alignment using low cost electronics

Hi All

I stumbled upon the following post on another UK forum.

"DIY laser leveling using webcam and laser level."
Link:

The poster provided a link to the software source code on Github
It creates a Gaussion graph of position, I tested the software using a crude test setup.

I obtained repeatable accuracy of 0.039mm (.0004" and I am sure that can be improved with a better setup....... remarkable.

Regards
John

There seems to be a rather large error in metric to imperial translation!

Andrew

That sort of technique can be made to work but there are many traps for the unwary. Was involved on and off with pixel based position measurement systems for most of may working life as a scientist / R&D engineer for MoD. Usually as the guy doing the grunt lab work to prove why a bright idea didn't work as it should.

Bottom line is that it starts getting very tricky once you go below pixel size. Remembering that the nominal pixel size is the spacing not the actual image sensor. Also remembering that simple gaussian plots to determine the beam centre hide a lot of assumptions concerning pixel sensitivity variations, beam quality and even diffraction effects in the sensor structure. A colour sensor will completely crap things up.

Probably reliable to a thou or three with care and thought but below that you are in for serious work. I tend to agree with the rule that once you go below a thou things get serious. Fast. I've no doubt that it can be done if you throw the processing power of modern desktop / laptop or even smart phone computers at it. But its very easy to be blinded by the data deluge and loose sight of the fundamental physics. Back in the day we never had the luxury of so much computer power so had to think very carefully about the physics.

Correct approach is to micro-scan the beam image over the sensor so everything can be properly defined. Beam needs to be at least 10 pixels wide. Shed loads of data but at least you know what you are dealing with.

In general interferometer methods are much easier as fringes are robust over small errors and easily observed by cameras. Just needs a good way of converting displacement to tilt.

Back in the day when I had to align off axis paraboloids I would set up a small He-Ne gas laser on each axis and zero out the fringes produced by back reflection on the front mirrors on both lasers. Thats four sets of fringes bought coincident at both ends. But I really, really needed a foot diameter diffraction limited beam!

Clive

Edited By Clive Foster on 24/08/2019 12:06:20

Hi Andrew

I meant to say this as posted at the other place

"Now for the exciting part; Hidden at the bottom of the Gaussian graph there is a tiny numeric value, It followed my indicator movement when I moved the slide. very roughly a value of 1 to 4 on the graph value to 1 to 4 ticks on the 0.01mm indicator. I assume this correlation may not have been planned however with the particular camera I used it was very convenient.

I ran the test several times and it was repeatable.
0.01mm is equal to .00039" (about 4 tenths)
A very good result from a crude setup."

There is an error in the scaled value that needs to be explained. The dial indicator readings and the values on the screen correlated and were repeatable.

Hi Clive

Yes there is more work to do on this project, For a start I want to use the sensor to work with backlit .008" stretched wire without a laser and no lens, just the shadowed area of the wire on the sensor. I am working on something I intend to show at the MSMEE Exhibition in Melbourne in 5 weeks.

You are right using a modern PC enables number crunching speed unobtainable not that many years ago, T 640x480 webcam came from the junk bin. Now 1080p camera modules are dirt cheap some under 20 dollars

The colour image is converted to grey-scale by the software. 

I will keep working on this project.

Regards
John

Edited By John McNamara on 24/08/2019 12:59:40

Edited By John McNamara on 24/08/2019 13:06:43

.

Very interesting project, John

You might see some useful improvement if you use a camera with an actual monochrome sensor.

MichaelG.

.

Yes, Frances ... that was my point

A monochrome sensor has one pixel per pixel [if you see what I mean]

A typical colour sensor [Bayer pattern] has four sub-pixels per pixel.

Gotta believe that the monochrome sensor will give crisper resolution.

MichaelG.

.

For those unfamiliar with the Bayer pattern ...

This is why (Red + Green + Blue) = 4

https://en.m.wikipedia.org/wiki/Bayer_filter

Edited By Michael Gilligan on 24/08/2019 16:59:19

John,

Skipping back to post #19 of that thread ...

Many thanks for the link to Oleg Borisov's thesis

MichaelG.

Anyone who has seen my posts on autoguiding a telescope won't be surprised!

Neil

Hi All

I Bumped this post because there are some worth wile changes.
**LINK**

There is an executable version available, A Jar file so you will need the current version of Java installed.
If you have a webcam preferably one with a lead attached not the one in your laptop although it will work. it is all you need to experiment.
**LINK**

Regards
John

PS:
The program code for the latest version has also been updated on Github for those that can program.

Edited By John McNamara on 27/08/2019 01:32:56

Thanks, John ... Much appreciated

MichaelG.

Good News, John

I downloaded the [zip file] executable to my MacBook Pro ... and was initially concerned to see 'laser.bat'

Fortunately however, the Mac seems able to launch the JAR file directly.

... All looking very impressive so far.

MichaelG.

.

I have sent you a PM, John

MichaelG.

Thanks Michael I responded with an email.

Hi All

I Just uploaded some further results from my experiments here: **LINK**

I am rather pleased with the results and there are further improvements possible.

Regards
John.

John,

I have sent another P.M.

[not specific to this discussion]

MichaelG.

Hi Michael
I sent you a Gmail and a PM

Hi All

I ran further tests with a new webcam, outstanding results!

This project keeps getting better.

See the end of this post here.

**LINK**

Regards
John

.

That's great news, John ... Keep up the good work

This looks like being a very effective spin-off from consumer technology !!

MichaelG.