Machsupport Forum
Mach Discussion => Video P*r*o*b*i*n*g => Topic started by: sshneider on April 19, 2008, 02:39:23 AM
-
Though I would start a new thread just for Clouds.
Here are a couple of some of my latest results. Both of these scans were done with both Y & X passes. Stitiching/striping courtesy of MachCloud! I wish I knew more about how to mesh accurately and/or add surfaces. I can see a lot of detail in my clouds but it would really be nice to see an accurrate fully 'surfaced' rendering. One must crawl before they walk.
Sid
-
Hello Sid,
> Thought I would start a new thread just for Clouds.
Great idea.
> Here are a couple of some of my latest results. Both of these scans were done
> with both Y & X passes. Stitiching/striping courtesy of MachCloud!
This looks like a great data set. Also appears properly proportioned. Scanned hands that have been shown in other threads have appeared to have extra long fingers.
Are the stray points from the blanket?
Can you show a photo of your camera and laser setup?
What camera and laser are you using?
What is the useful camera field of view and how far is the target from the camera and from the laser?
What is the laser diameter in the working area?
Did you take this data with the house lights on?
Try scanning a sphere and see if the cloud appears circular viewed from any direction.
Tom Hubin
thubin@earthlink.net
-
Hey Tom,
Getting right to the heart of the matter eh? ;)
This looks like a great data set. Also appears properly proportioned. Scanned hands that have been shown in other threads have appeared to have extra long fingers.
Awww Shucks but, I am but an instrument. 8)
Are the stray points from the blanket?
Yes there are. But, I orginally was using my wood table as a background and it was worse- hence the darker blanket. It helped reject the Zero plane points but I would like to find something better. So far my best results have been to raise my subject about 1" above the zero plane and then set the Ignore setting to .75". Works pretty good. That's how the Ark was done. The Frame was directly on the Blanket.
Can you show a photo of your camera and laser setup?
Sure, but probably not till Monday
What camera and laser are you using?
OOOOOOOOOh that's classified Area 51 security clearance stuff ;D ... Camera = Gigaware available at a Radioshack near you- I think I paid $19.95. It runs under the Logitect Driver so it's probably a knock-off. $4.00 Laser came from the Ebay listing someone posted on another thread in this forum- probably made by illegal, shoe-less child labor ::)
What is the useful camera field of view and how far is the target from the camera and from the laser?
OK, the height is easy for me to answer- around 10" (depending upon if you mean how high above the Zero Plane Vertically OR if you mean taking into account the angle of the camera and projecting a straight line from the lens to the Zero point of the laser. 10" is the diagonal dimension. I guess this is an Isoceles triangle geometry calculation?
I'm not sure what you mean by "USEFUL" FOV? I know what my setting is but, I don't think that's what you mean. How do I figure this out- sorry if this is a lame question.
What is the laser diameter in the working area?
Uh, another Techie question that I'm not sure what you mean. The laser is projecting a line on my table about 6" long. So I guess that would be a 12" Diameter- although I'm not sure why this is important to know.
Did you take this data with the house lights on?
Unless my cat turned on the lights at my home they were off. :D The lights at my shop were I did the scanning were also off (sorry I just couldn't resist) :P
Try scanning a sphere and see if the cloud appears circular viewed from any direction.
OK, I'll try it next week some time and let you know.
Regards,
Sid
-
Hello Sid,
> I'm not sure what you mean by "USEFUL" FOV?
The laser plane Y dimension and XZ dimension that the camera sees. You can use a ruler or graph paper in place of the laser plane and see that with the camera under normal lighting.
>> What is the laser diameter in the working area?
> Uh, another Techie question that I'm not sure what you mean.
The collimated laser is often 1mm to 2mm thick as it leaves the laser module. If there is an adjustable laser focus lens then you can adjust the laser to be skinnier at the target. A skinnier laser is more concentrated and therefore more immune to background noise (e.g. shop lights).
Is your laser collimating lens adjustable? If so, have you adjusted the lens so that the laser is skinniest near the center of the scanning range? Can you estimate the thin dimension of your laser line near the center of your target range?
> The lights at my shop where I did the scanning were off.
What happens if the shop lights are on? Can you show something scanned with shop lights off and again with shop lights on, just to see if the shop lights are problematic?
>> Try scanning a sphere and see if the cloud appears circular as viewed from any direction.
> OK, I'll try it next week some time and let you know.
A ping pong ball is spherical with a nice surface. A tennis ball or baseball would show well but these have irregularities in the surface. Still, they should appear circular from any direction.
Tom Hubin
thubin@earthlink.net.
-
Hey Tom,
The laser plane Y dimension and XZ dimension that the camera sees. You can use a ruler or graph paper in place of the laser plane and see that with the camera under normal lighting.
I dunno, I need to measure- I'll get back to you on this ok?
Is your laser collimating lens adjustable? If so, have you adjusted the lens so that the laser is skinniest near the center of the scanning range? Can you estimate the thin dimension of your laser line near the center of your target range?
Yes, Yes & I'd say approx 1mm.
What happens if the shop lights are on? Can you show something scanned with shop lights off and again with shop lights on, just to see if the shop lights are problematic?
OK, I will try this next week
A ping pong ball is spherical with a nice surface. A tennis ball or baseball would show well but these have irregularities in the surface. Still, they should appear circular from any direction.
Anyone know where to buy ping pong balls? ;)
Sid
P.S. I hope others will post some point clouds too!
-
Hello Sid,
> I hope others will post some point clouds too!
I wish I could. I have all the material and tool bits here now so hope to be able to build my laser triangulation gauge over the next few days. I will be attending a local amateur machinist club (http://www.cams-club.org/) meeting Thursday evening and would really like to show my laser triangulation gauge there.
After that I still have a system problem with Mach3. I have tight quarters in my apartment's small second bedroom shop. I prefer laptops to control my Sherline 5410 CNC mill and Sherline 4400 (soon to be CNC) lathe. I have been able to get Mach3 to run ok and my camera software to run ok and both to run ok simultaneously. I have also been able to get the camera to run within Mach3 for Video Window and Probing. However, Mach3 slows to a crawl, at about 1% normal speed, after either video plugin is loaded.
I have this situation on three different laptops. I may be forced to use a desktop plus keyboard plus monitor but, frankly, I have no idea where I will put all that stuff.
Tom Hubin
thubin@earthlink.net
-
Here is my first recognizable cloud.
It is a small key. Too small for a house key. Probably from a mailbox or large padlock.
The file was made with MachCloud. I tried both the November and December versions of MachCloud. I have had no luck yet with the meshing feature so I am just posting the cloud.
Tom Hubin
thubin@earthlink.net
-
Wow- I wish my first scan turned out that good! Nice!
Looks like you need to play with the "Ignore" variable to eliminate the Zero plane (or raise the key above your table a little bit).
Also, Meshing won't work unless the Cloud is sufficiently Decimated. You need to Decimate a few times to reduce the point cloud density then try Meshing.
HTH,
Sid
-
Hello Sid,
>> Wow- I wish my first scan turned out that good! Nice!
Well, I did qualify that by saying it was my first (recognizable) scan.
> Looks like you need to play with the "Ignore" variable to eliminate the Zero plane
> (or raise the key above your table a little bit).
I thought I could take that out with MachCloud. I was able to set data boundaries for positive and negative X and Y and for positive Z. Attempting to set for negative Z deleted all the data before I could mark the boundary. I would prefer to leave in some table points since thickness is a necessary characteristic of a key if you want to duplicate it or catalog it.
Of course, I am just playing with a key. There is no real goal here. However, one of my customers in the mid 1980s was interested in cataloging keys to create a data base for keys used in crimes.
Tom Hubin
thubin@earthlink.net
-
OK, here's my theory based on playing around thus far...
I have never been able to really make the Ignore variable work correctly when the scan subject was lying directly on the table. I DO calibrate my block to the table BUT it seems that whenever I increased the Ignor variable to a point that it would actually eliminate the zero plane, I would lose data from the subject.
This could be just my set up and the inaccuracies thereof. My work around was to elivate the scan subject by placing a shim underneath. Obviously the shim was smaller than the subject so it would not be scanned.
You can do some pruning in MachCloud but, it's hard to get that kind of accuracy due to the size/inability to zoom in (as of yet I hope ;D )
I have been able to do some 'precision' pruning using Rhino but we've already been there. Hopefully Art will be finished with LazyTurn soon and will resume work on the Video Probing package (wink wink-nod nod) ;)
I have another idea that I want to run by you but I'm short on time right now and it should be put under a different topic- TTYL!
Sid
-
Does anyone have an actual point cloud file they can send me? I have a program called Vrmesh that is specifically for creating meshes from point clouds. It a great program but a bit pricey.
Thanks,
Andre
-
Here ya go...
-
Here is the point cloud that goes with the mailbox key JPEG that I posted previously.
Tom Hubin
thubin@earthlink.net
-
Hello,
Here is a single pass scan of a boxend wrench. yStep is 0.010 inches. Speed is 5ipm.
The wrench measures about 5.5 inches but shows up in Rhino as closer to 5.0 inches.
Camera is a PlayStation 2 (silver color housing, not the black version) with exposure control set at 1.
Higherst exposure setting would be 130 and lowest setting would be zero.
Laser is a 5vdc, red (I forget which wavelength) module from Aixiz. The laser module is gettting power from a second USB port.
I did not get this on ebay since Aixiz does not list their 5vdc modules with adjustable lenses.
Computer is 800MHz Dell Inspiron 8000 laptop.
Tom Hubin
thubin@earthlink.nety
-
Hello,
Here is a single pass scan of a penny. yStep is 0.010 inches. Speed is 5ipm.
The penny is 0.75 inch diameter. This appears to be correct in the Y direction but more like 0.87 shown in the X direction.
I was guessing at the camera FOV. I will try to get the scaling right another time.
Camera is a Digi Watchport/V2 640x480 CCD webcam with exposure control set at about 1ms.
Exposure range is anywhere from 1/30000 sec to 1/4 sec.
Laser is a 5vdc, red (I forget which wavelength) module from Aixiz.
The laser module is getting power from a second USB port.
I did not get this on ebay since Aixiz does not list their 5vdc modules with adjustable lenses.
Computer is 800MHz Dell Inspiron 8000 laptop.
Attached Txt file is the point cloud produced by Mach3.
Attached Gif file was produced by Rhino.
Tom Hubin
thubin@earthlink.net
-
Hi Tom,
Very neat your penny scan. I would like to reproduce it and I have two questions.
1. I would love to see a print screen of the 3DVideoProbe panel to make the scan of the penny. Could you please post it ?
2. I have loaded the file thubin080712CloudPenny03.txt in MachCloud but when I go to mesh it the cloud disappears :o and no mesh is being done. Have you mesh that cloud?
If yes what settings do you make on the BallPivot mesh panel?
Thanks,
Gaston
-
Hello Gaston,
> 1. I would love to see a print screen of the 3DVideoProbe panel to make the scan of the penny.
> Could you please post it ?
I have posted a scan setup from today. The video shows a single line through the center of the penny. Exposure time was set at 1/3840 seconds and the resolution was 640x480. Mirror (left to right) and Flip (top to bottom) were selected since the camera is upside down.
I think that only the FOV is different. I changed it from 18.1 degrees to 15.4 degrees and got better proportions on the penny. This is about right for the +12mm focal length lens that I use on the Digi Watchport/V2 camera.
> 2. I have loaded the file thubin080712CloudPenny03.txt in MachCloud but when I go to mesh it the cloud disappears and no mesh is being done.
> Have you mesh that cloud? If yes what settings do you make on the BallPivot mesh panel?
I have had very little success with any of MachCloud's features. I have been using the free evaluation version of Rhino but have done little with it other than display and measure pointclouds.
I have figured out how to delete points from the cloud in Rhino. Never could get that to work right in MachCloud.
I could not get meshing or any of that kind of stuff working in MachCloud. I have not tried that in Rhino.
What camera are you using? Do you know the lens focal length? Do you know the field of view?
Tom Hubin
thubin@earthlink.net
-
Hi Tom,
>I have posted a scan setup from today. The video shows a single line through the center of the penny. Exposure time was set at 1/3840 seconds and the resolution was 640x480. Mirror (left to right) and Flip (top >to bottom) were selected since the camera is upside down.
Thank you, that was very useful. I do not have the facility to set the exposure time, the camera is taking 29.970 frames/sec and that's it.
>I think that only the FOV is different. I changed it from 18.1 degrees to 15.4 degrees and got better proportions on the penny. This is about right for the +12mm focal length lens that I use on the Digi Watchport/>V2 camera.
I have set the FOV to 70 which when calibrated with a 1" Styrofoam block give a distance approx right. The x-y dimensions seem also right.
>> 2. I have loaded the file thubin080712CloudPenny03.txt in MachCloud but when I go to mesh it the cloud disappears and no mesh is being done.
>> Have you mesh that cloud? If yes what settings do you make on the BallPivot mesh panel?
>I could not get meshing or any of that kind of stuff working in MachCloud. I have not tried that in Rhino.
It seems that if the dimensions of the object are too small (not the number of points in the cloud) meshing swallows it ???
>What camera are you using? Do you know the lens focal length? Do you know the field of view?
http://www.supercircuits.com/Security-Cameras/Board-Cameras/PC302XS with a digitizer The resulting image is 640x480 and black and white
I have found that the if I lower the voltage on the laser line generator I get a cleaner image of the line.
I have to work on my setup so I can change the distance between the camera and the laser easily. Right now the laser is pointing straight down and the camera is at ~45deg.
As for meshing in Machloud:
I do not understand what the Delauney Set is used for ???
BallPivot mesh seems to works ok with the Prop.txt supplied by Art but I have not so far succeed to pick winning values for the Ball Diameter, shortest Edge and Max Angle to produce the same kind of results on anything I have scan so far. I am using inches dimensions. I wish somebody could explain how to set those parameters?
Regards,
Gaston
-
Hello Gaston,
> I do not have the facility to set the exposure time, the camera is taking 29.970 frames/sec and that's it.
On the Video Probe screen, select Stop Video then select Source and/or Format to see if you have any exposure or resolution options. Then select Start Video to resume. Don't be surprised if Mach3 crashes when you try this.
You may find similar options with the camera vendor's software. Sometimes you can setup your camera options with the vendor's software and Mach3 will use that setup. I can send you stuff for other cameras that seems to be fairly generic.
> I have set the FOV to 70 which when calibrated with a 1" Styrofoam block give a distance approx right.
> The x-y dimensions seem also right.
I see that the camera is shipped with a lens for 70 degree FOV. The lens is replaceable. I suspect that the thread is M12-0.5mm. If so, you can get lenses on ebay. Search for M12* and lens* in title and description...or I can point you to some I have seen listed. I intend to get some of these M12-0.5mm threaded lenses from ebay but have not yet done so.
With a 70 degree field of view you will be able to look at large items in coarse detail. You will not get enough detail to pick out Lincoln's head on a penny. You will also have a good deal of distortion. You probably can determine by diameter that the coin is a penny, nickel, dime, quarter etc. To reproduce my results you will need a 12mm focal length lens or longer.
> > What camera are you using? Do you know the lens focal length? Do you know the field of view?
> http://www.supercircuits.com/Security-Cameras/Board-Cameras/PC302XS with a digitizer.
> The resulting image is 640x480 and black and white.
Black and white is good. Color info is not useful and results in larger data transfer that takes longer to transfer and process. Also, B&W cameras usually produce sharper images than color cameras.
> I have found that the if I lower the voltage on the laser line generator I get a cleaner image of the line.
That is probably due to saturation of the CCD array. The fix is to reduce the exposure time or sensitivity of the camera. When you reduce the laser power you also reduce the CCD saturation, so that helps, but this does not reduce background noise like reducing exposure time does.
> I have to work on my setup so I can change the distance between the camera and the laser easily.
> Right now the laser is pointing straight down and the camera is at ~45deg.
Neither the camera nor the lens should be pointed straight down. The camera and the lens need to be pointed 90 degrees apart. I have chosen to have the laser 30 degrees CW from the Z axis and the camera 60 degrees CCW from the Z axis, as shown here: http://www.machsupport.com/forum/index.php/topic,5023.msg42913.html#msg42913 .
> As for meshing in Machloud:
> I do not understand what the Delauney Set is used for ???
> BallPivot mesh seems to works ok with the Prop.txt supplied by Art but I have not so far succeed to pick winning values > for the Ball Diameter, shortest Edge and Max Angle to produce the same kind of results on anything I have scan so far. I > am using inches dimensions. I wish somebody could explain how to set those parameters?
Sorry but I am also clueless there.
Tom Hubin
thubin@earthlink.net
-
Hi Tom,
>Neither the camera nor the lens should be pointed straight down. The camera and the lens need to be pointed 90 degrees apart.
> I have chosen to have the laser 30 degrees CW from the Z axis and the camera 60 degrees CCW from the Z axis
As the same setup is used both for calibration and measurement why 90 deg be so important?
I see that the amplitude of the deviation would be maximized with 90 deg but as I understand it the results would not be more valid.
Gaston
-
Hello Gaston,
> > Neither the camera nor the lens should be pointed straight down. The camera and the lens need
> > to be pointed 90 degrees apart.
> > I have chosen to have the laser 30 degrees CW from the Z axis and the camera 60
> > degrees CCW from the Z axis.
> As the same setup is used both for calibration and measurement why 90 deg be so important?
> I see that the amplitude of the deviation would be maximized with 90 deg but as I understand
> it the results would not be more valid.
If you were using a pinhole camera the geometry would not matter.
However, when you use a conventional camera with a lens which is parallel to the detector,
the optimal layout is with the object also parallel to the lens and the detector.
With the object (i.e. the laser plane) at 45 degrees to the camera lens, the image of the laser plane will NOT be parallel to the lens and therefore will not be superimposed on the detector. That means that you cannot get the entire laser plane in focus. That will hurt your accuracy.
The smaller your field of view the more of a problem you will have if the laser plane is not parallel to the lens and detector planes. It is possible that the accuracy using a 70 degree FOV is so bad already that the angle the camera makes with the laser does not make much difference.
There is a way to point the laser straight down the Z axis and set the camera at an oblique angle smaller than 90 degrees but this requires that the detector be mounted separately from the lens so that the detector can be tilted at the appropriate angle in accordance with the Scheimpflug condition.
For more detail on the normal arrangement and the Scheimpflug arrangement please read some of my posts in other threads in the Video Probing forum.
Tom Hubin
thubin@earthlink.net
-
Hi Tom,
I see your point. Thank you for the explanation.
You have mentioned previously here:
>I see that the camera is shipped with a lens for 70 degree FOV. The lens is replaceable. I suspect that the thread is M12-0.5mm. If so, you can get lenses on ebay. Search for >M12* and lens* in title and description...or I can point you to some I have seen listed. I intend to get some of these M12-0.5mm threaded lenses from ebay but have not yet >done so.
My search on Ebay was not very productive so I think I could use some pointers here.
Thank you,
Gaston
-
Hello Gaston,
> > I see that the camera is shipped with a lens for 70 degree FOV. The lens is replaceable.
> > I suspect that the thread is M12-0.5mm. If so, you can get lenses on ebay.
> > Search for M12* and lens* in title and description...or I can point you to some I have seen listed.
> My search on Ebay was not very productive so I think I could use some pointers here.
Below is a link to an inexpensive assortment on ebay. This vendor has many M12-0.5mm thread lenses listed
on ebay and at his store. I have no affiliation with this vendor and have not purchased anything from him.
I would suggest that you choose the longer focal lengths since you probably already have the shortest one.
The longer focal lengths will give you a smaller FOV with greater magnification and detail.
http://cgi.ebay.com/Any-4-pcs-3-6-6-8-12-16mm-CCTV-board-Lens-set-dome_W0QQitemZ270255253150QQcmdZViewItem?hash=item270255253150&_trksid=p3286.m14.l1318
Tom Hubin
thubin@earthlink.net
-
Thanks Tom for the Ebay lens package reference,
I will look that up.
Right how I'm making a setup with 90deg (45-45) between camera and laser :)
Gaston
-
I was trashing some old pills and decided to try scanning bottles, caps and pills.
The closeup of the round pills shows impressed "I-2". I think these pills came from an Aleve bottle.
The closeup of the capsule shaped pill shows impressed "NUPRIN".
Text on the Nuprin bottle is readable because the camera exposure time was set so that the white areas produced a strong and usable signal while the dark areas produced a weak and unusable signal. The weak signals appear as gaps in the bottle's surface.
The video probe used has a PlaytStation 2 silver camera which has about 45 degree fov. The laser is 5mw, 5vdc red module from Aixiz directly (not ebay). USB cable provides power.
Attached are the point cloud TXT file and a GIF file of the Rhino screen showing the point cloud.
Tom Hubin
thubin@earthlink.net
-
Hi Tom
Nice to see how good your scans are getting.
I'm curious how we are going to get deeper profiles scanned. Can we step the Z down and take another slice then stitch the two meshes together or is getting a wider FOV setup the better way to go.
Cheers.
Mark
Still slowly building my CNC
-
> Hi Tom
> Nice to see how good your scans are getting.
Thanx.
> I'm curious how we are going to get deeper profiles scanned. Can we step the
> Z down and take another slice then stitch the two meshes together or is getting
> a wider FOV setup the better way to go.
See my suggested volume scanning at http://www.machsupport.com/forum/index.php/topic,7084.msg46309.html#msg46309
Small FOV is preferred for any detail and for less distortion. Then as many overlapping passes as necessary to get the job done.
Art, or whoever is working on the 3d video probe plugin, needs to set that up to be user friendly. Otherwise, you have to do each pass separately then combine later with a spread sheet to get the offsets right.
Tom Hubin
thubin@earthlink.net
-
Hello,
Here is a scan of a Quarter. Done in two overlapping passes since a quarter is a little wider than the video probe FOV.
Video probe camera is a 3Com Watchport/V2 set for 640x480 resolution and using a 12mm focal length lens.
Video probe laser module by Aixiz is 5vdc, 5mw, 635nm, 39 degree line lens, unknown collimating lens focal length. Arrangement of camera and laser shown in several threads over the last few months.
Scaling is close but not quite right.
I deleted the base points in Rhino so as to make the details of Washington's head more apparent. However, the data set in the text file does include the base points. That is one of the reasons that the text file is so large.
Tom Hubin
thubin@earthlink.net
-
Here is a two pass scan of a dead cricket done with a 12mm focal length lens on a 3Com Watchport/V2 webcam.
Maybe I can find a Praying Mantis for a live scan.;)
Tom Hubin
thubin@earthlink.net
-
Hi Tom,
You are becoming the expert at scanning small object :) You are now ready to machine one of these beast :)
What's next?
Have you tried to translate your data to mm and do a mesh in Machcloud?
Gaston