Designing new light, projector mount, and camera for table

The overhead was not as good. The picture is very sharp, but the barrel distortion is quite severe.

Direct from the camera:
1744159280097.png


OBS Filter applied:
1744159313778.png


After I saw this, I bought a bunch of different lenses on Amazon to see if one of them will be better. Some were labeled "anti-distortion", but we know until its tried, that might all be bs.

I also put my iPhone up there, and just like 5 years ago, the picture was perfect. But different than 5 years ago, Apple now has a feature called "continuity", where I can wirelessly connect an iPhone to the Mac and it will share its camera. I still don't want that hassle for a permanent installation, but I used that for streaming a pool match at the pool hall and it actually worked great. I'm almost certain if I dedicate an old iphone up on the ceiling, I'll have to climb up the table a lot to wake it up, etc.
 
Last edited:
Next up is this camera. It's similar to the prior one. A bit cheaper in price and build quality, also a bit smaller. Only has USB3 out, not HDMI out (which is fine with me). This one had a fixed 3.2mm lens, advertised as "no distortion".
1744162535988.png


1744162605985.png


The USB cord hit the ceiling, I'll probably get a right angle adapter if I end up keeping this.
1744162692198.png


I borrowed a mount from the end rail camera to clear the USB cord. Note, one of the biggest pains in setting these cameras up is getting them perfectly square to the table, and perfectly centered. The ball swivels make this quite difficult to do (and actually impossible is the swivel is not aligned with the center of the table). I may make a 3d printed bracket after I figure out which camera I'd like, that allows translation in addition to angle control.
1744163148544.png



Well what do you know, no distortion! The lines are all straight! The balls are smaller closer to the lens, and bigger furthest, but so far that's been true of all of these, including when I had the DSLR up there. This looks really good to me. Front runner by far thus far. The table is very slightly crooked due to the mount, which may affect the focus a hair.
1744162898744.png
 
Of note, I'm taking screenshots of the video output in OBS, and setting the pixels to 1000 wide for each table picture I'm pasting here. It's losing a bit of quality during that process. The balls all look a bit sharper in focus before I do that.

Also of note the 4k resolution of all of these new cameras is much better than the 1080P. I actually have the OBS set to 1080, so the feed from the camera gets scalled down. The end result is way better, because if I need to crop a picture in OBS to center the table perfectly, it will have a lot more headroom to do that. With the swivel mounts, the most important thing is to get the camera square to the table, even if it's not centered. Then in OBS, I can crop it to center it, or correct for any tilt. (Best of course is to center it with the mounts).

5 years ago the 4k stuff was expensive. Now its commodity.
 
Upon closer inspection of the Mokose 3.2mm picture, it looks like the long rails are very slightly curved inward. Also notice the balls furthest from the camera are biggest.
1744167719828.png


If I use the filter to add fisheye, it makes the balls in the center biggest:
1744168172949.png


Here is it with a correction applied to get the long rails straight. I think the balls are all closer in diameter compared to the first pic? I can't tell for sure.
1744168395046.png
 
Here is an interesting phenomenon I first noticed with the Nexigo 1080P webcam yesterday. I dismissed it at the time, but now I see it again. It seems color affects the distortion.

I have the CB and the 8 ball frozen and on donuts to ensure they don't move. This picture is from my phone held in my hand at the side of the table.
1744168922246.png


Now here it is on the overhead. The CB looks a hair bigger than the 8, right? Not a big deal since we know the balls change size as the distance to the camera changes.
1744168983734.png


However, if I reverse their positions, the CB is still bigger than the 8! How can this be?
1744169070876.png


Don't trust the video on pool streams, boys and girls;)

Of note, when I had the balls frozen on the Nexigo, the effect was much more severe. I already broke down that setup. I may put it back up later just to show this.
 
iusedtoberich said:
Here is an interesting phenomenon I first noticed with the Nexigo 1080P webcam yesterday. I dismissed it at the time, but now I see it again. It seems color affects the distortion.

I have the CB and the 8 ball frozen and on donuts to ensure they don't move. This picture is from my phone held in my hand at the side of the table.
View attachment 818067

Now here it is on the overhead. The CB looks a hair bigger than the 8, right? Not a big deal since we know the balls change size as the distance to the camera changes.
View attachment 818068

However, if I reverse their positions, the CB is still bigger than the 8! How can this be?
View attachment 818069

Don't trust the video on pool streams, boys and girls;)

Of note, when I had the balls frozen on the Nexigo, the effect was much more severe. I already broke down that setup. I may put it back up later just to show this.
Click to expand...
Are you sure its just not the color of the balls causing a size illusion??
 
Before I tear down the setup to try another camera, here is a couple min long video of the mokose 3.2mm with the +2.5 anti-fisheye filter applied. (the prior video was without the filter). Does the ball rolling look jitter to you boys? I can't tell.

 
Have you considered an action cam like a DJI osmo 5? Maybe a dumb suggestion, I know nothing about cameras. I use an osmo 5 for livestreaming and videos of my border collies catching the Frisbee. It's pretty good quality video.
 
Next up is trying different lenses on the 2 bodies I have.

First up I decided to try an expensive prime lens. Most of the lenses are in the $15-30 range, this one was $120. This brand also has 6mm, and 8mm prime lens (plus much more prime lenses, per the screenshot).
1744253305920.png


The lens does seem to be quite high quality. All metal, nice fit and finish, and the glass looks much bigger than the other lenses that came with the two bodies.
1744253223578.png


Here is the mokose 3.2mm prime on the left, and the "machine vision" 4mm prime on the right.
1744253516466.png



Unfortunately, the 4mm "machine vision" prime was a C mount, and the camera bodies were a CS mount. I bought an adapter ring, but it made the lens too far from the camera censor, and focus was not possible. I'm brand new to this mount system so don't know the ins and outs. The lines do look straight, however. Here is the 4mm prime on the Makose body.
1744238782148.png


I was able to get it to focus better by jerry rigging the adapter. This shows that 4mm is not wide enough. The lines are straight.
1744239074813.png


Next I wanted to see if the lens or the body was correcting the barrel distortion. I put the ELP zoom lens on the Mokose body. The distortion was bad. This proves the lens, and not the body, is both the culprit and solution.
1744240774540.png



Here is the ELP body with the Mokose lens. I had the same issue that the lens could not focus because it was hitting the cmos sensor. But, the lines are straight, so it shows the ELP body is ok, and the ELP zoom lens was the culprit of the distortion.
1744241576739.png


Here is the ELP body with the expensive 4mm prime lens. The adapter worked properly on this combination. Focus is great, lines are straight. This lens might end up working well on the table end rail view. Ill try it later there.
1744241921039.png
 
oneball said:
Have you considered an action cam like a DJI osmo 5? Maybe a dumb suggestion, I know nothing about cameras. I use an osmo 5 for livestreaming and videos of my border collies catching the Frisbee. It's pretty good quality video.
Click to expand...
I had looked into the go-pro's a couple years ago. Also a buddy has one he uses at the pool hall at times. They were not quite what I was looking for.
 
iusedtoberich said:
Here is an interesting phenomenon I first noticed with the Nexigo 1080P webcam yesterday. I dismissed it at the time, but now I see it again. It seems color affects the distortion.

I have the CB and the 8 ball frozen and on donuts to ensure they don't move. This picture is from my phone held in my hand at the side of the table.
View attachment 818067

Now here it is on the overhead. The CB looks a hair bigger than the 8, right? Not a big deal since we know the balls change size as the distance to the camera changes.
View attachment 818068

However, if I reverse their positions, the CB is still bigger than the 8! How can this be?
View attachment 818069

Don't trust the video on pool streams, boys and girls;)

Of note, when I had the balls frozen on the Nexigo, the effect was much more severe. I already broke down that setup. I may put it back up later just to show this.
Click to expand...
White makes stuff look fat - even to a camera it appears.
 
iusedtoberich said:
Before I tear down the setup to try another camera, here is a couple min long video of the mokose 3.2mm with the +2.5 anti-fisheye filter applied. (the prior video was without the filter). Does the ball rolling look jitter to you boys? I can't tell.

Click to expand...
To me the cue ball looked a little jittery on some shots compared to the vid in post #104. I thought the video in post 104 looked fine. Actually either one looks fine to me. I probably wouldn't notice much difference unless I watched an hour long match on each.
 
Next up is this 3.5-8mm zoom lens on the ELP body. This lens was dirt cheap.
1744686428183.png


1744686505203.png


1744686726189.png


This lens is a no-go for overhead. Lines are very crooked, and focus was not possible on all portions of the table. I do think a lot of the miss-focus might have to do with the camera not being perpendicular to the table. It is extremely difficult to adjust it to the correct angle, especially with these wide angle lenses.
 
Epiphany or phony from the GC release:

If you had a reflector system (previously known as mirrors) on the lights, you could shoot directly at 'em or at least collect multiple undistorted images. (?)
 
To improve the camera placement to the table, I went looking at what was out there, and came across these devices made for surveying. They are called Tribrachs. They have trickled down to camera use and amazon pricing and I ordered one. I'll have it in a few days and try it. I was actually thinking of designing and 3D printing something, but if this is off-the-shelf and works, screw that;)

1744688533273.png
 
You must log in or register to reply here.
Back
Top