iRVision and user fraims

    External Content www.youtube.com
    Content embedded from external sources will not be displayed without your consent.
    Through the activation of external content, you agree that personal data may be transferred to third party platforms. We have provided more information on this in our privacy policy.

    I'm trying to setup iRVision a user frame following the above video however on my On my R-30iA I only have [Trhee Point] [Six Point(XZ)] [Six Point(XY)] [Two Point + Z] and [Direct Entry] I do not see the option to do a four point calibration. When I try to do a three point calibration off the dots like in the video it tells me the points I selected are invalid. :help:

    When calibrating the frame it seems to be in the Tool frame and not the User frame. I have tried setting the frame in tp programs and by pressing 'COORD' button (both with and without shift) but the MENU -> SETUP -> FRAMES still is stuck on tool frame and not user frame. I have had it in user frame before I just don't know how I get back to it.

    In my R-301A V7.40 iRVision Documentation apendex 1 page 143 it seems to have instructions for calibrating using the tool frame 6 point calibration method (see attached picture of my reverse engineered documentation). I'm just not understanding what "manually enter: W = W + 90" means, where is this entered? Alternatively does anyone know how to get it into four point user frame calibration mode instead?

    I cant seem to find the other option. I was however able to see that if I open dual displays and I open frames twice the second one is in user mode. I'm still really freaking confused but I got it to go. But after power cycling I cant seem to use the move to command to go to recorded points, I'm also having problems going to points from the vision system, I get a MOTN-018 error, which gives this excellent troubleshooting tip.
    "Cause: Position not reachable Or near by singularity
    Remedy: Reteach the position that is not reachable"

    I think my two problems are connected but I don't know what I'm doing wrong :angry_face:

    Not sure what I'm missing here but in the left panel view of your images under Tool frame there is an 8 but under the coord key you have selected 9. Both images top and bottom show tool frame 8 but looks like there is a mismatch between what is identified and what is selected but I could be wrong. Are you using tool frame 8 and user frame 9? Or are you trying to make them both frames number 9?

    Quote from rafahil

    What are you trying to do?
    A userframe?
    Or Toolframe?

    is your calibration grid mounted on the robot? or is it fixed?

    Camera and calibration grids are not mounted to the robot arm. Pretty sure I want user frame. I just need to be able to move to and pick up found items reliably, in the end I don't really care about the method used I just need to reach this objective.

    Quote from Jayrohr62


    Not sure what I'm missing here but in the left panel view of your images under Tool frame there is an 8 but under the coord key you have selected 9. Both images top and bottom show tool frame 8 but looks like there is a mismatch between what is identified and what is selected but I could be wrong. Are you using tool frame 8 and user frame 9? Or are you trying to make them both frames number 9?

    Hmm, I also seem to be confused about what I am doing here with tool frames, perhaps this confusion is the cause of much of my problems.

    Quote from HawkME

    Menu > IrVision > utility > grid frame set

    Are you doing a fixed mount or robot mounted camera?
    Fixed

    I was able to find Menu > Utilities > iRvision > Automatic Grid Frame Set

    Is this the method for automatically calibrating the frame by mounting some sort of identifier on the robot arm and having the robot jog around the image area?

    If so it may be more dificult for me to use this, my robot is the M-1iA (the delta spider robot) and the camera is somewhat obscured when the robot arm is over the area I am interested in. I could create a calibration tool for it to grasp that sticks out in front and somehow do an offset from that but it seemed to me to be more hassle then just using the calibration grid and then manually calibrating the robot to that grid. So I clearly am goofing something up but I cant for the life of me figure out what it is.

    So I re did everything, calibrated the user frame, calibrated the camera and made a new test program. Program finds the part, robot moves but not anywhere close to the part, its also changing in the roll direction (should only change in xy and maybe z). I made sure everything was on the same user and tool frames for all of it.

    Edited once, last by FrankM (December 15, 2016 at 1:05 AM).

    Update: I was able to figure out how to change frame types in Setup > Frames, you get the "type" "detail" "other" "clear" "setind" option if you simply press > "PREV"

    I had not even realized I was stuck in a sub option that pressing "PREV" would return me out of.

    I do still very much have something wrong with my calibration, when I try to move to a vision offset the robot moves unexpectedly and does not get me to the part I want to grab.

    The simplest method for calibrating a camera is the grid method. 1st, place a cal grid in the field of view of the camera. Ideally it should take up most of the FOV. 2nd, teach a temporary tool that is a pointer. 3rd, teach the grid as a user frame (3 or 4pt), preferably a uframe you don't use (ie 9). 4th, calibrate the camera with a grid cal type. DO NOT MOVE THE GRID BETWEEN STEPS 3 and 4 OR YOU WILL HAVE TO START OVER. The entire process takes about 5-10 minutes once you're used to it. There are other methods but if this is your first foray in to vision, use the KISS method until you have a good understanding of it. Hopefully this helps.

    The simplest method for calibrating a camera is the grid method.
    I hope so but so far I've just been bashing my head against the wall. I see mentioned in passing methods to calibrate by grasping the calibration grid with the robot or creating a series of points using a cross hair attachment. Perhaps I should try one of these methods? My manual is rather sparse on them.

    1st, place a cal grid in the field of view of the camera. Ideally it should take up most of the FOV.
    Done, unfortunately due to the physical structure where I am using the camera it must be off to the side a little or it would be off in the z axis by several inches. I understand that areas outside this grid view are considered to be indeterminately calibrated but the robot seems to be unable to navigate to even parts well within the derided area.

    Some related questions: Is the sequence of the numbering important, they do not seem to be labeled in the order I would expect. Are the lens focal distance and position of camera values important, the calibration does not seem to be auto generating them very well. What is this reference position, do I need to do anything with it?

    2nd, teach a temporary tool that is a pointer.
    Luckily my end effecter comes quite nearly to a point and can grasp very fine things, I have been using the actual tool frame calibration, there is no need to use a temporary one instead is there?

    3rd, teach the grid as a user frame (3 or 4pt), preferably a uframe you don't use (ie 9). 4th, calibrate the camera with a grid cal type. DO NOT MOVE THE GRID BETWEEN STEPS 3 and 4 OR YOU WILL HAVE TO START OVER. The entire process takes about 5-10 minutes once you're used to it. There are other methods but if this is your first foray in to vision, use the KISS method until you have a good understanding of it. Hopefully this helps.

    I believe I am doing all of this but somehow everything is going to crap, usually when I try to move to a part the robot wont even move at all erroring out saying the position is out of reach. If it dos move it does not move anywhere near where it should. Very frustrating.

    I think I see one problem, in your grid calibration you've set you application frame to the same frame as the grid but your code shows uframe 9. The application frame is the frame you'll be working in, and the . Usually is 0 (world). In your 2d vision process the offset frame will be the grid frame # and the application frame is what frame you're working in. I probably should've gone in greater detail earlier. Unless you require a uframe for what you're doing, try teaching your pick in world (uframe 0) as you'll eliminate the possibility of error on one side of the process. This way your accuracy will only be affected by utool an cal grid and not an additional uframe. Look at these pics from a previous setup and hopefully it will help.

    Quote from rumblefish


    I think I see one problem, in your grid calibration you've set you application frame to the same frame as the grid but your code shows uframe 9. The application frame is the frame you'll be working in, and the . Usually is 0 (world). In your 2d vision process the offset frame will be the grid frame # and the application frame is what frame you're working in. I probably should've gone in greater detail earlier. Unless you require a uframe for what you're doing, try teaching your pick in world (uframe 0) as you'll eliminate the possibility of error on one side of the process. This way your accuracy will only be affected by utool an cal grid and not an additional uframe. Look at these pics from a previous setup and hopefully it will help.

    Thank you, I'm actually starting to get somewhere now. The robot is now moving in relation to the part (though offset and its rotating the wrist 90°) but at least when I move the part up 1" and to the right 2" the robot arm moves an appropriate distance. But its still not moving to the part. I still don't quite understand what exactly the difference between "Application Frame: User Frame:" and :Calibration Grid Frame: User Frame:" is or what the reference data section of Vision Tools configuration does.

    Edited once, last by FrankM (December 16, 2016 at 4:56 PM).

    So once you've calibrated the camera with grid cal, you'll need to teach the GPM tool in your vision process. Once you've taught the find(gpm), the next step is setting the reference data. This creates a "golden part" or reference used when calculating the offset. In the most basic sense, it's a point taught on a user frame. That user frame is manipulated(moved) by the vision process and the point therefore moves with the frame. Now, once you've set your reference DONT MOVE THE PART UNTIL YOU'VE TAUGHT THE ROBOT PICK POS. The reference point is ZERO. To teach your pick points, 1st Run Find (again), 2nd GET OFFS and 3rd teach your points. Any point using the VReg Offs will ask if you want to subtract the offset data from the position, click YES. All that's left is to test the points. Long term, you really should look in to taking the 2D Vis class.

    Quote from rumblefish

    ... To teach your pick points, 1st Run Find (again), 2nd GET OFFS and 3rd teach your points. Any point using the VReg Offs will ask if you want to subtract the offset data from the position, click YES. All that's left is to test the points. Long term, you really should look in to taking the 2D Vis class.

    Thank you so much, it is working now. I was not teaching the parts off the found location, that was a step I was missing all along. The accuracy is not the greatest but it is at least getting close each time and grabbing it 70% of the time. Is there a way to teach on parts at multiple locations?

    Fixing lighting and getting a lens that is more zoomed in are things on the list. Are polarizing filters recommended?

    Edited once, last by FrankM (December 19, 2016 at 8:39 PM).

    You are not going to get accuracy with that calibration. Its best to do a 2 plane calibration. Also, you have it set to Orthogonal. It needs to be set to calculate the focal distance. An orthogonal calibration is the least accurate and shouldn't be used. Finally, you need to set the z height from the offset frame to the top of the part in your vision process tool. The z height is critical to accuracy.


    Sent from my VS985 4G using Tapatalk

    Edited once, last by HawkME (December 17, 2016 at 10:44 PM).

    Quote from HawkME

    You are not going to get accuracy with that calibration. Its best to do a 2 plane calibration. Also, you have it set to Orthogonal. It needs to be set to calculate the focal distance. An orthogonal calibration is the least accurate and shouldn't be used. Finally, you need to set the z height from the offset frame to the top of the part in your vision process tool. The z height is critical to accuracy.

    I have the M1iA 0.5 and was not able to mount the cameras exactly perpendicular to the pick areas due to the struts dangling down to the wrist. I could zoom out a on the cameras and make them perpendicular but I would loose resolution on the rather small parts I'm trying to see.

    Is it best to A) mark it in software as perpendicular even though its off by a little, B) zoom out and make it perpendicular C) Leave it as is.

    *Pictured robot is not mine but is the same model.*

    First a correction on what I told you earlier. Since you are using a fixed camera and fixed grid you will need to select a 1 plane calibration, Projecting=Perspective, Override focal distance=yes, Dist=focal length of your lens. (In order to do a 2-plane calibration you would need to mount a target on the EOAT. This works very well if you can do it, & no manual touch up required!)

    I try to get my grids perpendicular to the camera axis, but depending on the accuracy you need, it may ok, you just need to test it. If you are worried about loosing resolution by moving the camera farther away you can always go with a longer focal length lens which will "zoom in" on the image. How far off from perpendicular are you and what lens are you using?

    I would redo the calibration this way, leaving the camera where it is and see how good your accuracy is before changing the camera position. Make sure to set your Z height in the vision process tool!

Advertising from our partners