Camera setup and alignment

[CoreyMac (Corey McCormick)]

Thank you very much for the quick response. I will answer inline…

@berainlb

@CoreyMac

I think it would help if there was a note in the onscreen instructions that has the download link to tell you to print it out on 1/2 of a sheet of paper and what the dot dimensions should be.

While the dimensions for the pattern are not in the calibration wizard itself, the dimensions are listed on the documentation and Oz calls out that it’s roughly a quarter of a letter size paper in the video included in the documentation.

I think you were answering while I was still editing the post. Man you are fast!

@berainlb

Note that the precise size of the pattern is irrelevant, however. It’s the relative size of the pattern to the captured image that’s more important. So ideally the following would be met:

1. The pattern size fills roughly 1/9 the field of view at the focal distance expected between lens and work material.
2. You’ve focused the lens well, again for the focal distance expected
3. You have extremely even lighting for the pattern
4. You have a very blank background, ideally lightly colored. You want to maximize contrast to the test pattern. You most definitely do not want anything in the background that in any way resembles dark circles.

I did all of this.
The lens is fixed focus and clean and looks normal in Windows using the builtin camera app and in the thumbnail window in LB.
The full sheet of paper with dots is about 1/9th of the area of the laser bed but it was tried quite a few times on both full sized 8.5x11 and 1/2 size as the documentation indicated. Both plain white paper and white card stock were tried and the pages were very flat and laying on top of a single sheet of box grade tan cardboard that was long and wide enough to basically cover the entire bed to use as a blank background to eliminate any visual clutter.
I used another piece about 18x18inches for the 1,2,3,4 target alignment that I could lay on top of this, and it is about 1/8" thick.

@berainlb

@CoreyMac

the scale is off by 50% and nothing lines up.

This is a large enough change to indicate that there’s a fundamental problem with how it’s working. Can you confirm that the focal distance from lens to material is unchanging? The alignment will literally only work properly at one focal distance.

Are you doing camera aligning to freshly burned targets that have been unmoved from burn to capture?

I only printed a new one of these if the distances between the camera and the bed changed for that alignment.

@berainlb

When you do your capture during camera alignment, can you confirm that the image is distortion free? If it’s not then that means there’s something wrong with the initial lens calibration and that needs to be corrected first.

Mostly, but the very edges of the bed image tend to have some aspect distortion no matter what else is tried.

@berainlb

@CoreyMac

Fisheye or normal also does not seem to help

If your lens has any distortion at all you would need to choose fisheye.

It has some and I have tried both ways

@berainlb

@CoreyMac

The images at 2M are currently so poor that I would not think even 8M images would be massively clearer

In what way are they poor? Just in terms of resolvable image? Can you confirm that you’ve focused the lens?

I suspect 8MP will be certainly better than 2MP but beyond that will be diminishing returns for a bed of your size.

What are you expectations for placement accuracy? I have an 8MP camera on a 700mm wide bed and I’ve found I can get to roughly sub millimeter placement but not better. While resolution could be better I suspect it’s flaws in the distortion correction and alignment process that limiting the accuracy before resolution limits. That could potentially flip in very large beds but uncertain.

I would not expect it to be amazing, just clear enough to know that things are working correctly. Things like:

I can see the engraved lines from the laser (they are about .25 - .5 mm wide most of the time) and know how precisely things are being placed/engraved without go to the laser to look directly. I was going to put it inside an enclosure and wanted to be able to make multiple cuts without opening the enclosure between every cut to watch the framing run live. I can use Windows camera app to watch it move, but the actual laser tracing path is hard to see that way and you can’t really line up multiple cuts that way… I wanted enough resolution to tell when a piece that has been cut drops down a bit as it is released so I know if a another cut pass is needed on a particular cut without jostling things around first.

This is an example that might help illustrate.

The blue outline is the actual LB object and is about 37mm wide in LB. The cutout actually made is also 37mm wide save for the kerf so the actual laser calibration seems fine, but this was where the last calibration of the camera left things. The camera is pretty much above the center of the bed, but the image is zoomed about 200% in when I show the overlay.

When the one on the lower right was actually cut, this is the LB screen. The blue outline needed to be about 2 inches below and a little to the side for that lowest cut and the scale is pretty far off as you can see. When I framed the cut, it put the frame right where the cut really happened, it was just the image that didn’t match.

I have tried a couple of dozen calibrations over the past month or so and more more I learned, the better they got, but this scale issue I have not eliminated yet. I am fine with going through everything again, I was just trying to figure out what I was not doing correctly so I didn’t repeat the mistakes. I am sure this is supposed to work, I just haven’t found the right dance steps.

Sometimes it inverts the image between two sample images like this whether I use either of the camera interface options.

Sometimes not… You can see that things are focused and clear.

I will go and try again, though I am not really sure what to change. For example, should I use a full sized dot sheet or a 1/2 sized sheet? A full sized sheet is ~1/9th of the area per the documentation you mentioned as #1 above, but the correct 1/2 sized sheet is far less even though the documentation says it should be small. With a larger bed of 46x81cm you can’t really meet both criteria at the same time can you?

To avoid any rotation issue requirements, I have also moved the camera mount so that the native camera angle is correct. I can adjust the vertical height some, but it was not clear exactly what should be in the frame. The laser gantry and print head, the entire bed and rails, only the printable area, etc… I would have some control over how the image frames, but not perfect. It’s aspect ratio and field size versus the bed laser area and frame are fixed but the camera position can be moved. It is on a 2020 aluminum extrusion riser above the bed.