DIY Laser levelling using webcam and laser level

**devmonkey** · 18-08-2019

Originally Posted by John McNamara

I agree that you could use the output from the sensor to directly plot position.
Using a simple webcam with a 2mm sensor to centre a .008" piano wire or laser, the effective measurement range would be maybe 1mm. The active area of the sensor needs to be found in some way, unless the wire or beam is positioned within the active area all the time.

Hi John,

With a self levelling laser hitting a 2mm square target over the entire machine shouldn't be too hard assuming you have levelling feet on the machine, we all do this or better when we put up a shelf. This assumes we can weld the frame to within 2mm.

Should also be noted that it is only the vga (640x480) web cams that have this small sensor area I think these are 1/6" diagonal sensors, other similarly priced but higher resolution cams have larger sensor areas. For example the raspberry pi camera (1/4") has a 4.6mm diagonal and another sensor I have the OV2640 has 4.5mm. It seems that the sensors are classified by the diagonal size of the sensor of which the active area with the pixels is smaller, this is why 1/4" > 4.6mm.

I like the idea of levelling a jig then grinding the surface as an alternative to shimming. What sort of grinder would you use for this?

I can imagine a jig short jig say 500mm that clamps around a piece of box section that can be micro adjusted for twist and level. I we bolted aluminium to the top of the box section you could probably skim it with a hand router with a small fly cutter in it. Skim the section under the jig then move the jig along, re-align it with the laser/wire then repeat. When you finish one rail then move the assembly to the other without moving the reference laser to bring the other rail into plane. Maybe a wood power planer could be used for the skimming, would be scary though.

The jig itself needs to be planar as you say but this should be straight forward on a milling machine.

What are you thinking for differentiating twist from height variation? I think with the laser two image sensors stuck out each side could detect twist same when you tram a mill with two dial indicators on a T bar in the spindle.

Cheers, Joe

**John McNamara** · 18-08-2019

Originally Posted by devmonkey

Hi John,

"With a self levelling laser hitting a 2mm square target over the entire machine shouldn't be too hard assuming you have levelling feet on the machine, we all do this or better when we put up a shelf. This assumes we can weld the frame to within 2mm."

Assuming a vertical arrangement the only alignment required is to align the rough frame he two counter weighted wires. When a work piece is clamped in position it is also aligned to the wires.

The concept of using a "metrology frame" in this case for one axis is not new.
https://www.google.com/search?q=metr...5sCCs1M:&vet=1

It is important that the alignment wires are in line with the tip of the cutter at the point of cutting. This will greatly reduce any error caused by any angular misalignment of the z axis to the work piece. Abbes principal must always be honored.
https://www.google.com/search?rlz=1C...71.zasns7Fk5YU

------
"Should also be noted that it is only the vga (640x480) web cams that have this small sensor area I think these are 1/6" diagonal sensors, other similarly priced but higher resolution cams have larger sensor areas. For example the raspberry pi camera (1/4") has a 4.6mm diagonal and another sensor I have the OV2640 has 4.5mm. It seems that the sensors are classified by the diagonal size of the sensor of which the active area with the pixels is smaller, this is why 1/4" > 4.6mm."

I think sarting with a low cost solution first then if necessary upgrading to a more expensive camera.

------
"I like the idea of levelling a jig then grinding the surface as an alternative to shimming. What sort of grinder would you use for this?"

I had been reflecting on grinding but had substituted milling, strangely I should have written milling!
Grinders have a lot of inertia and take a long time to stop, they also make a lot of noise. A Small motor driven milling cutter ideally around 3mm can be fed relatively a lot slower and the motor can stop quickly. I suspect having a running spindle near the wire alignment system may cause the wires to vibrate. For high accuracy results it may be necessary to stop the spindle before each alignment cycle.

------
I can imagine a jig short jig say 500mm that clamps around a piece of box section that can be micro adjusted for twist and level. I we bolted aluminium to the top of the box section you could probably skim it with a hand router with a small fly cutter in it. Skim the section under the jig then move the jig along, re-align it with the laser/wire then repeat. When you finish one rail then move the assembly to the other without moving the reference laser to bring the other rail into plane. Maybe a wood power planer could be used for the skimming, would be scary though.

The jig itself needs to be planar as you say but this should be straight forward on a milling machine.

------
"What are you thinking for differentiating twist from height variation? I think with the laser two image sensors stuck out each side could detect twist same when you tram a mill with two dial indicators on a T bar in the spindle."

Two weighted wires suspended vertically will have no twist between the at all. These two wires are our meteorology frame. if we position our cutting stage relative to the wires using two cameras, make a peck cut then move X or Y,, check the cameras are still perfectly centered, make a new cut and so on our accuracy will be determined by the accuracy achieved by the camera system.

No system is without error however if you consider the above scenario the surface generated will have a surface that comprises a series of physical levels some high and some low. representing the accuracy of the measurement cameras and the positioning system.

Using a vertical alignment system, the exciting part is that the wires represent a plane that is far flatter than an inspection grade surface plate. The final result will be a surface that is extremely straight overall. With a surface similar to a scraped bearing surface.

Upon reflection I will make my test setup vertical.

As you may have guessed I have been thinking on this problem for a long time.

Cheers, Joe

Answers above.

**devmonkey** · 19-08-2019

Kitwn, about using it for tramming I would attach my sensor block to a T-bar in the spindle as if I was tramming a mill with a single dial indicator. Rotate the spindle and take sensor height readings at 0, 120 and 240 degrees rotation. Take plane described by these 3 3D points, calculate the plane normal and determine the error to the plane normal of your x axis plane, then shim the z axis until the normals are the same.

I will have a think about squaring but this is usually easy to do with dual x axis machines after skimming the bed by drilling 4 corners of a large square and checking diagonals using a stick with an indicator on the end, adjust home proximity sensors, re-home, repeat.

If at any point after initially bringing the x rails into plane you need to re-establish this plane in order to do something like tramming, it is unlikely the laser can be put back where it was initially. However since the rails are still planar the laser can be setup again and 3 points measured 2 from one rail, one from the other and the x plane normal established relative to the new laser normal. This is similar to moving between machine and workpiece coordinate systems in a CAM programme. For tramming we would align the spindle normal to the x plane normal via the laser normal, i.e. When the two errors are the same the spindle is trammed.

**Kitwn** · 19-08-2019

I'm very interested to look at your software on github. I'm no great programmer but have an unused Raspeberry Pi and a Pi camera which my brother -in-law sent me. I've been trying to work out what to do with them so this might be a possibility.This is exactly the sort of application he'd enjoy seeing it used for. It would be a large sensor block though with the Pi included as the camera attaches directly to it.
-
Part of my concern with working out an exact series of measurements is separating the possible sources of a given error and being able to align the machine in a systematic way. There's no point making the axis of the spindle exactly vertical until after you've ensured the axis along which the Z assembly moves is vertical and that can't be done until the Y rails and gantry mountings are set up and so on.
-
Some assumptions might be possible if your mechanical construction is very accurate but my machine is built using welded steel construction, a drill press and hand tools without access to a milling machine or other precision machining, so it's accuracy will only ever depend on being able to make measurements and correct errors with shims. I designed the whole thing with that requirement in mind but it's being able to make accurate measurements without spending more on a variety of gauges than on the machine itself that has made this thread so interesting. Like many hobbyists I have more spare time than spare money and there's great satisfaction in following Arthur Wellesley's maxim.
-
Kit

**John McNamara** · 20-08-2019

Hi Devmonkey

I downloaded and set up the camera software from Github.
setting up is quite a process, a mate who is a far better programmer than me led me through the environment variables setup. I am not used to this sort of programming environment.
Once installed correctly the software compiled without an issue.

I was only using the webcam on my laptop and a hand held laser line (not point) generator. It worked very well.

A logitech webcam was also tried it worked but there were strange fresnel like circles produced with the laser. The logitech camera had a quite large lens. The laptop camera did not display the same aberrations. its camera is tiny.

I also tried a fuzzy pencil line drawn on paper that worked OK too.
A piece of wire should be no problem.

Over the weekend when I have time I will set up a test on a granite surface plate with some precision gauges. and a proper camera with and without a lens on a stand.

It will be very interesting to get a better idea of repeatable accuracy. I am sure it will be very good.

Congratulations Joe, great job.

Regards
John

PS:
I have started to draw up the flat surface generator I described,
I want to make it 90% from Laser cut 5mm steel plate. Using a single universal beam as the main frame (100ub 23) I know that that profile is available in Australia and the UK Hopefully Europe and the US too. It will hold an up to 100x100mm RHS Workpiece. To be used as linear rail support beams for routers and mills.
I hope it works! There is only one way to find out.

**devmonkey** · 20-08-2019

Originally Posted by John McNamara

Hi Devmonkey

A logitech webcam was also tried it worked but there were strange fresnel like circles produced with the laser. The logitech camera had a quite large lens. The laptop camera did not display the same aberrations. its camera is tiny.

I also tried a fuzzy pencil line drawn on paper that worked OK too.
A piece of wire should be no problem.

If you are using a laser with the (crap) optics of a webcam then you will get all sorts of weird patterns and probably more problematic non-linearity meaning each pixel will register slightly different height errors, best to get an old or cheap webcam and rip the sensor out and bin the optics. If you want to stick with optics then use a piece of white paper as a screen between the laser and the camera, focus the camera on the screen.

With the pencil line did the gaussian detection work? I would be surprised if it did as the pencil line would be lower intensity than the paper. I would need to invert the intensity I would think for this to work properly with a pencil line or wire. Also to use as a shadow camera you will need to flood the background with an even light source, maybe a white led behind a piece of diffusion perspex (the translucent white stuff). This is to eliminate false triggering from the otherwise random background intensity of your workshop wherever the camera is pointing.

**John McNamara** · 20-08-2019

Hi Devmonkey

Yes the line did produce a gaussian plot. It was a crude test. I just placed the paper in front of the Logitec camera about 150mm away no backlighting just ambient fluorescent (Its night time) I did get the curve plot. it was all hand held testing lots of movement. I really want to test it properly. It will be interesting to try inverting greyscale?.

Cheers
John

**devmonkey** · 24-08-2019

Hi John,

Yes the tiny number is the deviation in pixels from where you pressed zero.

I don't think it was ambient light affecting your setup, the laser saturates the sensor. It is so sensitive it will easily pick up deflections in your floor from walking about. This happened when I was testing too, I had to sit completely still. It is less sensitive to walking about on a 6 inch bare concrete slab but you can still move it a few pixels.

You don't need a better camera you just need to rip the optics off that one. The difference in quality between webcams of the same resolution is simply the optics.

**John McNamara** · 24-08-2019

Hi DevMonkey

I just did this calc in Excel

sensor size in mm 2
/ pixels 640
Space between pixel lines = 0.003125 mm

I got roughly 1 integer value change to 0.01mm div on the Mitutoyo indicator.
Quite a big variation?

Regards
John

**devmonkey** · 24-08-2019

My setup works out at the predicted resolution of 3um per pixel. I don't think you can trust this at all with your setup as the optics are still attached but with nothing to focus on. You are also getting a load of noise on the gaussian, this is probably either a function of the crappy optics on the camera or the laser line is using a diffraction pattern rather than a cylindrical lens.

Rip the lens off the camera, then you will also be able to see if the sensor is indeed 2mm.

If you think about it, without optics and the laser projected directly onto the sensor moving the sensor 1 pixel will shift the maximum intensity one pixel, there is no other option as the sensor and beam have moved relatively and the intensity profile of a cross section of the laser line hitting the sensor has not changed.