machine vision

    Hi sir

    I am working on a new project which is detecting some object with an industrial camera and send its coordinates to the industrial robots (KUKA...), technically I will transform the object coordinates from pixel (taken with the camera) to world coordinates (robot coordinates)

    this my first project and really I need help

    my question is which are the best cameras for this kind of vision and what are the cameras that can be integrated with the industrial robots

    also, is the robot's controller can process images?

    Best regards

    Well that's a bit vague to give any reasonable answer.

    What kind of object are you trying to locate?

    How much data do you need from your picture 2D/3D

    What kind of accuracy are you expecting.

    Basically the only thing i can tell you is that, No the robot controller does not process images. But a industrial camera can send coordinates to a robot controller.

    So first figure out what you need, after that you can think about how you are connecting it to the robot controller.

    Every problem has a solution, that isn't the problem. The problem is the solution.

    ok but can you explain why? i see no speed, no accuracy, no palletizing...

    in fact not sure why they even needed robot, vision etc.

    could just let parts fall off conveyor directly into receiving container. :face_with_tongue:

    perhaps use counting scale to confirm quantity and stop conveyor...

    am i missing something?

    1) read pinned topic: READ FIRST...

    2) if you have an issue with robot, post question in the correct forum section... do NOT contact me directly

    3) read 1 and 2

    The camera will detect a shiny metal object like the picture below with dimensions (150mm x 50mm).

    And I just need the coordinates of the object to send it to the robot

    Quote from panic mode

    ok but can you explain why? i see no speed, no accuracy, no palletizing...

    in fact not sure why they even needed robot, vision etc.

    could just let parts fall off conveyor directly into receiving container. :face_with_tongue:

    perhaps use counting scale to confirm quantity and stop conveyor...

    am i missing something?

    As I said The camera will detect a shiny metal object like the picture with dimensions (150mm x 50mm).

    The conveyor will stop when the object enters the FOV of the camera till the robot picks it up.

    no important accuracy needed here just need coordinates

    I am working on this project as an end of year project and if this project went well the company will apply it to its different field (they try to automate everything there ):grinning_face_with_smiling_eyes:

    Hi Makrem,

    Consider the following:

    > 2D / 3D

    > Static mounted or robot mounted camera.

    > Lens : Cognex do a good lens calculator. FoV / DoV )

    > Lighting : you mention a shiny object, think about LED colour , diffused lighting.

    > Conveyor : If applicable, a MATT surface with obsurb lighting to give a more defined silouettee of the PatMax of the part(s) to pick / place.

    Fieldbus : If using a SMART CAMERA, you'll need the relevant fieldbus tech package most probably ( unless you already have this ). - Mainly applies if tethering directly to the robot.

    If using a GigE camera, if a Beckhoff PLC is your "Master", you can feed via EtherCAT and this has its benefits ( no BIT / BYTE / WORD swapping required), else Kuka VisionTech is my obvious choice as this is configured and controlled through the KONI port of the robot itself ( Kuka Optional Network Interface ).

    Also, PerceptionTech : If wanting a dynamic solution, Perceptiontech is just pure brilliance, We use this on the KRC4 platform, you load a CAD model into the "Camera", and the system detects and "Picks". - Draw backs are that the model generations can only been uploaded via the OEM ( RoboCeption)...They are working on a customer self suffiecient .EXE, but they for now offer a package of a number of model generations for a one off fee. ( NDA approved )

    This has it's benefits with KCP commissioning and a "Production screen" within the GUI.

    what will drive the above really comes down to £££££££.

    A good calibration artefact is a FIDUCIAL plate, sized for your resolution and accuracy requirements. Available from most vision suppliers.

    Repeatable and accurate vision systems often come down to good planning, this in my experience is lighting.

Advertising from our partners