Machsupport Forum
Mach Discussion => Video P*r*o*b*i*n*g => Topic started by: ART on November 13, 2007, 01:35:53 PM
-
Hi Guys:
This will be the first release of the video prober.I cant do a video on it, my system chokes hard trying to use video and capture
video at the same time. SO heres the interface running, Ill explain how it works. This is pretty early stuff, but likely to take a long time
to complete, so you may as well play with it and see what you can do in terms of point clouds.. who knows, you may have fun..
Heres the interface running..
-
Youll notice two windows on the interfce. The first is the actual video. Youll notice its dark. Thats because the lights are off.
Having lights on makes it very hard to reject specular noise. Youll also notice there is a raised section to the laser line. Thats beacuse there
is a 52mm block of wood on the table. This is the second step to starting up. When YOU start, have nothing in the field. Rotate your laser do the red line in the second
image is as straigth as possible. Then insert the block off wood or obejct to calibrate with., I use a simpel wooden block measured at 52mm high and 52mm wide. width doesnt really
matter, but you want to make sure the line below the block is showing well as well as the line on the block.
The slider between the images is a threshhold. Slide it up and down until you have what looks like a square wave on the line monitor window.
The White line incidently is raw data, the red line is CenterOFMass corrected, and centroided to produce a cleaner line. Lasers can create noisy lines.. so the red line is the
cleaned line. Leave all the numbers alone except for the Cube size, enter the size of your block there. (inches or mm's shoudlnt matter, whatever mode Mach3 is in..)
Then slide the threshold slider for a good square wave so it seems to be sensitive to the edges of the block but with as little noise as you can get, but make sure the
width of the block appears to be seen. Then hit calibrate. Repeatedly hitting calibrate, you should see now the laser angle showing and perhaps varying about .1 degrees, it
may vary more, ( or less) depending on the camera, and the laser. Not too important in organic scanning, but the more accurate is it, the better.. The camera height should now
show pretty close to how high the cameras focal plane actually is..
On the point control, you can see its set to reject any Z's less than 1mm, this just gets rid of the zero plane, I find it annoying..
Set a scan Y value of the length of your object to be scanned, you can remove the block anytime after cal'ing, but dont change Z height after calibrating..
Pressing Scan will now generate a point cloud called points.txt inthe mach3 folder. Youll find rhino has no trouble loading it. The demo of rhino is good for
25 saves, so as long as you dont save, you can view all the point clouds you like..
The samples setting is a averaging setting, the higher it is, the longer the scan will take. The Y stepover is the amount the camera will move on each scan.
Play, have fun, post any cool clouds here so we can all see them.. Ill post updates as I continue to make things more robust,
I can tell you laser probing is more of an Art than a Science.. ( no pun intended. :) )
Ive made quite a few cool scans lately, much work remains to be done to clean things up before I start to try meshing it all to a
3d model file, but I think a few woudl like to play with the clouds, and its all pretty easy to get used to really..
God luck,
Have fun,
Heres the plugin..
Art
-
One other note:
This thing is pretty CPU intensive, so performance will vary. :)
Art
-
Sweet! I'll give it a shot in the next day or so. The laser line is perpendicular to the vector of travel right?
- Alex O.
-
>>>>>>>LATER Art noted on theYahoo site this is a known situation and happens now and then... they are looking for a better OCX driver. <<<<<<<<
Hi Folks,
Always one to try new stuff, I fired up the new video prob(e)ing plug in and it brings up an issue I had quite some time ago with the video windows and Mach3. This is several versions later of the video window and at least one new computer and several new hard drives and XP installs. (Should be rid of my gremlins of old and have new ones by now). I played with it a bit and here is what I get.
If you use either video plug in all works just fine UNTIL you stop recording or exit either plug in. Then try to use the video window again –either one of them- and it pegs the CPU (dual cores) and nothing much else happens. Running this on my indoor machine with no driver loaded, not the machine on my mill out in the shop. You can exit and re open either plug in and they will not work. You must exit Mach3 to get things started again.
What you do get sooner or later is a little window saying “VideOCX(VIDEO) – Disconnect failed”. I will also leave this in the Mach support site and try and attach the little window.
(http://macmurchie.com/ocx.jpg)
Anyone have an idea what I am doing wrong? I am sure this must work just fine for others or we would hear more moaning and groaning. <Grin>
-Joemac ‹(•¿•)›
-
I tried the new plugin and had a couple of problems.
Firstly, the scan appears to be doing what its supposed to but upon completion and returning to 0,0, the points.txt file is 0 bytes long. I've tried several calibration settings and different scan items but points.txt never gets written.
Secondly, the calibration isn't returning expected results. The calibration reports 2.45 inches from camera to zero Z whereas the actual distance is closer to 4". The angle appears to calculate correctly though. I even went so far as to calculate the FOV as being 47.26 degrees but the Z distance comes back too short. Perhaps my camera is too close to Z Zero?
-
Hi:
When you go to calibrate, make sure you slide the threshold upwards till you can see the entire width of the calibration cube. Distance to the camera may be a bit off , I sometimes get the same thing. Ive working on that , but the only ill effect is registration of the striped data if you scan in the x dimention, I only scan in the Y at the moment.
If the points cloud is empty, try changing the "Ignore points less than n in Z" to -100 and see if it then posts points. If the nagle gets computed backwards, then its possible your points are upside down..
To compute the depth, the system finds a point on the upper point of the cube, and makes a 3d plane equal to the angle of the laser and uses that as an intersection plane.. since Ive only used it here, I cant verify it works on all lase rplanes, but Im curious what you get if you allow any point to be set. If it scans, then the only thing that woudl stop points coming out is the Ignore Points setting.. Let me know what you get..
Art
-
Art,
This is so awesome. I've got to build one of these.
I've read about other scanners that spin the object, are you telling me you just scan in the one axis? What I mean is, do you translate in one axis, or multiple axis?
Thanks,
Dave
-
Dave:
For this version at least, it scans only a linear section. When its all working Ill add a rotary scanner, but
its a challenge to get it linear only as yet. The idea is to allow you to copy a large carved door or something that lies on the
table in multiple stipes of aquisition. Lots of hurdles to cross, but its an incremental process..
Art
-
Success! (mostly). I set the "Ignore points less than n in Z" to -100 and the points.txt file saved. The scan itself looks kind'a squooshed in the X direction, which is probably due to the cam-to-Z0 distance being a bit off, and alot of noise. I'm going to tweak some stuff and see if I can improve it. BRB. :D
-
well, at least I can tell its a skeleton.. :)
Narrow your FOV, it may be too high, dont worry about height, just set it to 50 degrees, Im curious if that stops the stretching or smooshing..
Not too bad for a first attempt though.. at least its recognisable.. :)
Art
-
By the way, what did you process that to a model with, rhino?
Art
-
I used Rhino. Import points -> Points to Mesh -> Render -> Save as jpeg.
This scan looks way better. Moved the camera and laser up about 3", made a new exactly cubical 1.5"x1.5"x1.5" calibration block with a gray matte material attached to the top for better visibility and regularity, and turned the lights off during the scan. I'm going to experiment with photographic filters tomorrow night to see if I can leave the lights on by filtering out all colors but laser red (650nm).
Nice job on the plugin Art. It definately has great potential!
Edit: Oh yeah, and the calibration distance was within an 1/8" this time.
-
This plugin will be the most popular yet. It will save hours of time.
-
Hi Art
Is the video probeing doing to work on 2D part to get a toolpath just asking I am new to that one But it sure would work nice if it is going to do that
klmark