Enhancing Electronic Card Labeling Precision Using iRVision Fanuc

    Hello,

    I am currently working on developing an application for labeling electronic cards using a suction cup gripper mounted on my robot equipped with an iRVision 2D camera. There are always two markers present on the electronic card blanks, each marker being a 3mmx3mm square. My aim is to utilize these markers to precisely adjust the positioning of labels on the electronic card blanks. I deposit the labels using the suction cup mounted on the gripper.

    As a card blank can contain multiple cards, I employ an offset matrix to determine the necessary shift for each label. The electronic card blanks are fed through a supply system, and the referencing may sometimes shift due to variations in blank sizes. Hence, I am exploring the most suitable vision method for this process. In summary, I aim to correct the referencing offsets by capturing images of the markers using the iRVision camera. I have noted various methods such as Tool Offset, Fixed Frame Offset, etc.

    Currently, no testing has commenced, but I seek clarification on these points before initiating experiments. Additionally, I wonder if it is necessary to create a UFRAME from the supply system for this process.

    Thank you in advance for your assistance!

  • Go to Best Answer

    Is the camera mounted to the robot or fixed?

    Are the markers you are taking a picture of on a working surface or held by the robot when you take the picture?

    The answers to those questions determine if you are offsetting the frame or tool.

    A user frame is part of any vision process to determine the vision calibration grid location.

    Quote from HawkME

    Is the camera mounted to the robot or fixed?

    Are the markers you are taking a picture of on a working surface or held by the robot when you take the picture?

    The answers to those questions determine if you are offsetting the frame or tool.

    A user frame is part of any vision process to determine the vision calibration grid location.

    The camera is mounted on the robot

    The markers that are photographed are located on each electronic card blank, in the two opposite corners of the card (2 markers per flank).

    These markers are not fixed and must be photographed again each time a new PCB arrives via the destacker system.

    -> Process: The unstacker supplies and presents the PCB to be labelled > the robot takes a photo of the 2 markers on this PCB > and adapts its PR offset according to the situation.

    I've included a schematic diagram to make it easier to understand

    Furthermore, I don't know how accurate the label application will be (I know I can't tell without testing, of course). I've seen several vision methods in iRVision > GPM Locator Tool

    I've seen several vision methods in iRVision > GPM Locator Tool > Given the wide range of blank sizes to be processed, I won't be able to create a reference UFRAME per blank based on the markers. So I'm looking for a method that will allow me to use the information retrieved from the marker positions to adapt my first label application position (which has a direct influence on my subsequent PR offset).

    My idea was :

    I create a UFRAME on the unstacker system that holds the pcb blanks.

    And from this UFRAME learned with the 4 point method I shift the original system according to the position of the markers recovered by the camera and I establish a UFRAME on the unstacker system which receives the pcb blanks.

    • Best Answer

    So, the PCB blanks are (somewhat) randomly placed as they come to the robot? And the blanks come in different sizes? Meaning that the markers are different distances from each other when you run a different blank size?

    Are the markers reliably repeatable on each blank, relative to the blank and the PCBs on the blank? That is, if you have 100 blanks of Type A, are the markers reliable across all of them?

    What is your planned working distance for the camera? Do you expect to get the entire blank in the FOV, find both markers, and generate your UFrame offset? Or were you planning to zoom in, move the camera to each marker one at a time? For the latter option, you'll need some way to ensure that the markers fall within the FOV of the camera, even with the worst case placement error.

    Will the robot be aware of what size blank is coming? irVision multi-shot should handle this for you, but you would have to program different reference positions for each blank size. Multi-shot should generate a Fixed Frame Offset that works, without needing a UFrame for each blank.

    Completely agree with SkyeFire, you need to understand that detail and be able to reliably detect the markers. It would be easier if there was one unique marker in a single corner you could always find. A rectangular marker may be ok, but there are better options, such as an off center cross-hair shape.


    In addition, you would use a fixed frame offset. The vision calibration grid would be placed on the working surface and you calibrate and teach a UF for the vision grid there. The grid frame is all that is needed for User frames.

    After that you you use GPM locator and voffset to offset your label placement positions.

    Z height of the blank markers from that grid user frame must be a fixed known value.

    Hello !

    Update :

    - I've used for my tests the multi-view vision process of IrVision Fanuc. This method gave to me the possibility to automatise the calibration by 3 differents camera views (3 > visions RUN FIND).

    - In a second time i've used the VISION GET_OFFSET to update my CONDITION VOFSSET on my differents passage points.

    - Thoses methods works and this calibration offer more flexibility

    ! If you make an application with loop and offsets on points like me, pay attention to the order of the point terminations:

    For example >

    AND DON'T FORGET CALIBRATION IS KEY ! :upside_down_face:

    Thanks for your reactivity !

    MR.

Advertising from our partners