Hello Guest it is March 28, 2024, 03:57:16 PM

Author Topic: New 3d Video Probe  (Read 41501 times)

0 Members and 1 Guest are viewing this topic.

New 3d Video Probe
« on: March 23, 2009, 11:47:05 PM »
I have posted my first 3d Video Probe plugin. Calling it Scan3d.dll. No documentation yet. Iit should show the camera view.

http://www.machsupport.com/forum/index.php/topic,10958.new.html#new

Tom Hubin
thubin@earthlink.net

Offline cfell

*
  •  27 27
    • View Profile
Re: New 3d Video Probe
« Reply #1 on: March 26, 2009, 12:38:09 PM »
Tom
I downloaded the plug in and running it with a Logitech Pro 9000  I am very impressed with the resolution and sensitivity.  I attended your demo at Cabin Fever but lack most of the basics to connect any of the dots and get anywhere further with this technology.  Please keep going and I hope we get enough users interacting I can learn what I need to scan and produce an image.
Charlie
Re: New 3d Video Probe
« Reply #2 on: March 26, 2009, 01:22:12 PM »
Hello Charlie,

> I downloaded the plug in and running it with a Logitech Pro 9000  I am very
> impressed with the resolution and sensitivity.

Experiment with the Format and Source buttons to change the camera setup. The display images show as 320x240 resolution but the calculations are done with whatever resolution the camera is set for.

Experiment with image rotation and mirror. Rotation of zero, 90 CW, 90 CCW, and 180 degrees covers the 4 most likely ways to mount a camera. Mirror is only needed if there is an odd number of reflections in the mirror path. Most likely you will have no mirrors in the camera path so leave the mirror option unchecked.

Rotation of 180 degrees is needed if the camera is mounted upside down like my Watchport/V2 is.

You can experiment with scanning the three axes but the output into the file Scan3d.xyz will be strange until you calibrate. Ii have to leave shortly so will try to write about calibration tonight.

Tom Hubin
thubin@earthlink.net
Tom Hubin
thubin@earthlink.net 
Re: New 3d Video Probe
« Reply #3 on: May 26, 2009, 09:15:25 PM »
Hello,

I just uploaded the latest version of my 3d video probe plugin called Scan3d.

"http://www.machsupport.com/forum/index.php/topic,10958.msg74209.html#msg74209"

I included the source code with an "Open Source" license for those who wish to explore writing programs for accessing webcams.

Testing is somewhat limited right now since I have been away from home and my shop for the last 6 weeks.

Tom Hubin
thubin@earthlink.net

Offline Greolt

*
  •  956 956
    • View Profile
Re: New 3d Video Probe
« Reply #4 on: May 27, 2009, 06:01:42 PM »
This looks interesting Tom.

Thanks very much for sharing your work.

What sort of line laser is needed or recommended?

This is one cheap one I found.  Same as the cross one that I presently use for job setup.

http://www.dealextreme.com/details.dx/sku.5928

Thanks,  Greg
Re: New 3d Video Probe
« Reply #5 on: May 27, 2009, 06:30:40 PM »
This looks interesting Tom.

Thanks very much for sharing your work.

What sort of line laser is needed or recommended?

This is one cheap one I found.  Same as the cross one that I presently use for job setup.

http://www.dealextreme.com/details.dx/sku.5928

Thanks,  Greg

Hello Greg,

Mine is similar, almost identical. See link below for more detail.

http://www.machsupport.com/forum/index.php/topic,4989.msg50288.html#msg50288

Tom Hubin
thubin@earthlink.net

Offline Greolt

*
  •  956 956
    • View Profile
Re: New 3d Video Probe
« Reply #6 on: May 27, 2009, 08:00:49 PM »
OK I ordered a couple of those.  Not much to lose if they're no good.  ;)

Now what should I be looking for in a camera?

Resolution?  Compression? 

I see a lot of the webcams on ebay claim up to 8MP but in the specs are only 640 x 480

Greg
Re: New 3d Video Probe
« Reply #7 on: May 27, 2009, 10:59:00 PM »
OK I ordered a couple of those.  Not much to lose if they're no good.  ;)

Now what should I be looking for in a camera?

Resolution?  Compression?  

I see a lot of the webcams on ebay claim up to 8MP but in the specs are only 640 x 480

Greg

I am using 640x480 cameras. One is the Play Station camera. There are two silver models and one black model. All seem to work ok but the silver ones are easier to mount with a flat top surface. I drilled and tapped the top surface to have the same mounting pattern as the Watchport/V series of cameras.

The Watchport/V2 camera is great but seldom turn up on ebay. Can be bought new for $125 to $200. Can get on ebay for $50 to $100 when it turns up about once every 2 or 3 months.

The Watchport/V camera is interchangeable with the Watchport/V2 but much cheaper. It is from the 1990s so does not do compression so is slower. But it works fine on many machines. Can be bought on ebay for $10 to $30 including shipping. Often the IR Blocking Filter is hazy but it is easy to remove or replace. I have a few that the factory gave me so can mail you one if need be.

Most webcams can be used but may be a pain to mount. The Watchport webcams have 1/4-20 thread on the bottom so can be mounted on most tripods. There are two 1/8 inch diameter holes on the bottom that take 1/8 inch diameter dowel pins for consistent registration.

Take a look at some of my  hardware and drawings scattered around the Mach3 Video Probing forum.

Tom Hubin
thubin@earthlink.net
Re: New 3d Video Probe
« Reply #8 on: July 29, 2009, 04:33:07 AM »
Hello Tom,

I've been trying your new Scan3D plugin and have some questions about the calibration.

I used Art's video probe with some success earlier, but as your Scan3D has more possibilities in the setup of camera and laser - and in the end a better scan - I'm trying to figure out how to make it work.

If I understand right the direction of the laserline can be chosen according to ones own needs. For me this would be parallel to X-axis ( I tried parallel to Y-axis as well, problems are the same). I have both the camera and the laser pointing down at 45 degrees so the camera is looking perpendicularly at the line (in the middle of its vertical field of view).

I have a nice focusable line laser and Watchport V3 camera with a set of different lenses (tried 8mm F2.0 and 12mm F2.0). I have as well a Logitech QuickCam Pro. The Watchport has a problem of a bigger distortion of the picture (a straight line seen as slightly bent). The QuickCam has a zoom that demands a calibration every time you plug it in after having disconnected it for milling work.

With Art's videoprobe the quickCam gave ok results but took always a lot of time setting up - the Watchport was useless. With your Scan3D I understood the calibration should take care of - at least some of - the distortion. And what's nice no calibration needed every time before a scan.

The problems:

1) I can't get the calibration process - done as you describe in the short manual - to correct the "bending". This is easy to see by scanning a flat surface with two parallel passes resulting to two "trays" with edges not meeting.

2)As I understand the calibration should automatically correct any slight deviation in the direction of the camera in relation to the laserline. I still get  the edges of two adjacent passes not meeting but as two slightly inclined bands. (This is easier to see using the QuickCam as the scan is not "bent")

3)The whole of these problems may be a result of some of the calibration process missing... In the Scan3D window after calibration the fields for FOWs, Elevation and Azimuth all show "1". ?

Hope you can make give me some advice how to solve this since your Plugin is a big step from the previuos one.

Thanks a lot.
Eero
Re: New 3d Video Probe
« Reply #9 on: July 29, 2009, 06:38:09 AM »
Hello Eero,

> If I understand right the direction of the laserline can be chosen
> according to ones own needs.

Yes, that was the plan. More accurately the camera plus laser line assembly can be oriented as you wish. The relationship between the camera and the laser is fixed. The laser plane must be perpendicular to the camera lens axis.

> For me this would be parallel to X-axis

That is the way Art arranged his. Mine should work there as well.

> ( I tried parallel to Y-axis as well, problems are the same).

I presume that you calibrate for whichever axis that you are using.

> I have both the camera and the laser pointing down at 45 degrees
> so the camera is looking perpendicularly at the line (in the middle of
> its vertical field of view).

Sounds ok but I would like to see a photo or a drawing that shows the camera, the laser module, and the laser line. Can you post photos or links here for the benefit of others?

> I have a nice focusable line laser and Watchport V3 camera with a
> set of different lenses (tried 8mm F2.0 and 12mm F2.0). I have as
> well a Logitech QuickCam Pro. The Watchport has a problem of a
> bigger distortion of the picture (a straight line seen as slightly bent).

The longer focal length will usually have less distortion. Smaller field of view means more magnification for more detail too. Can you identify the lenses or the source? Are they part of the set that is sold for the Watchport or did you maybe find them on ebay.

> The QuickCam has a zoom that demands a calibration every time you
> plug it in after having disconnected it for milling work.

I can see where that would be very inconvenient.

> With Art's videoprobe the quickCam gave ok results but took always a lot of time setting up - the Watchport was useless.

I was able to use the Watchport/V2 with Art's plugin. I don't have a V3 to test with but I would expect it to work ok. I have one out for repair so may know more when it returns.

> With your Scan3D I understood the calibration should take care of -
> at least some of - the distortion.

That is the plan but not tested a lot. It seems to work on my setup but I use a lens that seems to have little distortion anyway. The math can be made to fix worse distortion. I just need to see some bad results first.
 
> And what's nice no calibration needed every time before a scan.

That was just a matter of saving the cal data. I think Art just did not get around to doing that when he was tinkering.

The problems:

1) I can't get the calibration process - done as you describe in the
> short manual - to correct the "bending". This is easy to see by
> scanning a flat surface with two parallel passes resulting to two
> "trays" with edges not meeting.

Sounds like the calibration may not have been done. The defaults should produce something recognizable but probably do little to correct distortion. I would like to see the data files and photos of the scan setup.

2)As I understand the calibration should automatically correct any slight
> deviation in the direction of the camera in relation to the laserline. I still
> get  the edges of two adjacent passes not meeting but as two slightly
> inclined bands. (This is easier to see using the QuickCam as the scan is not "bent")

I am not following you. I need to see the setup and the results.

3)The whole of these problems may be a result of some of the calibration
> process missing... In the Scan3D window after calibration the fields for
> FOVs, Elevation and Azimuth all show "1". ?

I think those are default values. You should get something like 45 degrees for elevation for your setup. It is possible that the point of the needle is not being seen or that the calibration scan not working right.

You need to be sure that the needle point is detected by the camera when the laser hits the point. That should be the only thing visible during the cal scan. You probably need to set the camera exposure time to a small value, like 1ms, so that the field is black except for a small area where the laser hits the point of the calibration needle. 

I am guessing by your numbers that you are using mm. I tested with inches but that should not matter.

Is the camera upside down or right side up or sideways? Photos of your hardware off the machine and on the machine would tell me a lot.

The camera orientation needs to be selected so that the line is horizontal in the second display. Ideally, that view should appear right side up for the part but would probably work upside down too. The line should not appear vertical in the second view.

Tom Hubin
thubin@earthlink.net