Feature recognition inside iRVision

    Hi

    I have a Roboguide cell configured with iRvision that looks something like this picture :

    As you can see there are some parts stacked on top of each other on a pallet and on one of the top one theres a red part inserted. What I want to do is to use the camera mounted on the robot tooling to grab a picture of the top parts and inside this picture have iRVision detect the red part. The idea would be to detect the red part makes inside the picture and use its position inside a program, heres what it looks like from the camera pov :

    Im new to iRVision and I've been unable to figure out how to do this so I would appreciate some help!

    Thanks

    Edited once, last by chrysis (September 24, 2024 at 4:39 PM).

  • SkyeFire September 24, 2024 at 3:25 PM

    Approved the thread.

    That is a very low F# lens. I would recommend against that, as you will have not the greatest guidance in the real world. Lots of fish eye.

    A higher up, fixed mount camera is typically used in depalletizing operations.

    Anyway, assuming your camera is calibrated, you could teach a simple 2d vision find with a GPM to find the red part. You would then train a reference position, and use that in your pick process.

    Check out the Fanuc position converter I wrote here! Now open source!

    Check out my example Fanuc Ethernet/IP Explicit Messaging program here!

    Thanks for the input!

    Yes I realise that the lens is low but I wanted to keep it on the robot as much as possible! However if its not possible I will mount it higher up.

    Also I understand how to find it with a GPM but if I understand correctly the GPM returns a position for the part in pixels (and degrees). How do you go about converting that value to cartesian coordinates?

    I'm also exploring the option of using a 3d camera to facilitate the process, do you think it might be easier? Currently Im having difficulty teaching the 3DL plane command tool. I've added carboard cuttouts in between the stacks to help.

    Thanks

    Quote from chrysis

    ...

    Also I understand how to find it with a GPM but if I understand correctly the GPM returns a position for the part in pixels (and degrees). How do you go about converting that value to cartesian coordinates?

    ...

    For an application where you want to use vision to locate a workpiece and then move the robot to the correct position for pick up, then this is the procedure:

    1. Train the workpiece with a GPM locator (for instance)
    2. Select Offset Data Calculation Tool and perform a Snap & Find*
    3. Set the reference position for your vision process.

    4. (in your TP program) Touchup your pick position (WITHOUT moving the workpiece from it's position when the ref.pos in step 3 was set)
    5. Run the vision process (VP) and get the vision offsets (Get_Offset)
    6. Use the VOFFS, VR[...] motion option in your motion instructions involving the pick position (pre-pick, actual pick, post-pick)

    The system will use all available information that you provided during setup such as:
    *Camera Calibration
    *Camera location at the moment of image snap
    *Offset fram specified in your VP
    *Ref.pos in your VP
    *Taught pick pos.

    and will then automagically offset the frame (a temporary/virtual frame) so that your taught pick position makes the robot move to the found position.

    You don't need to make any calculations of your own :smiling_face:

Advertising from our partners