Gas Leak Detection with Image Processing

    • New

    Hello everyone,

    I'm currently working on a project that involves controlling a Mitsubishi industrial robot (RV-2FD) using a Raspberry Pi 4B, and I would really appreciate your insights and suggestions.

    Project Overview:

    The aim of my project is to perform visual inspection on copper pipes used in systems like air conditioners, dryers, and refrigerators — specifically at their welding joints, where cooling gases (like freon) are used. I want to detect possible defects or leakage points using image processing, and then guide the Mitsubishi robot to those coordinates for further inspection or marking.

    Here’s how I plan to approach it:

    • I'm using OpenCV on Raspberry Pi for real-time image processing.
    • A camera module attached to the Raspberry Pi will capture images of the copper pipe welds.
    • I’m creating a custom dataset of defective and non-defective weld joints to train a model for detection.
    • The image will be processed on the Raspberry Pi, and defect coordinates will be extracted from the image.
    • These coordinates need to be translated into robot motion commands so the Mitsubishi robot can move to the detected defect points.

    Communication & Integration Questions:

    1. What is the best way to establish communication between the Raspberry Pi and the Mitsubishi robot? Are there supported protocols like serial (RS-232), Ethernet/IP, CC-Link, Modbus TCP, or any others that are compatible with both?
    2. Would it be better to use a PLC as an intermediary, or is direct communication between the Raspberry Pi and the robot feasible?
    3. Are there software tools, SDKs, or APIs available for Mitsubishi robots that can be integrated with Raspberry Pi (Linux-based OS)?
    4. How can I send coordinates from the Raspberry Pi to the robot controller? Is there a standard for converting image coordinates to robot world coordinates?
    5. Can the robot receive commands from external systems (e.g., via socket communication, HTTP, or a serial port)?
    6. Are there any existing examples or libraries (open-source or vendor-provided) that could help bridge Raspberry Pi with Mitsubishi robots?

    Ultimately, I want to make the Raspberry Pi act as both the vision system and the decision-making unit, while the Mitsubishi robot executes the movements based on the coordinates extracted from image analysis.

    Any documentation, tutorials, experience-based suggestions, or links to similar projects would be extremely helpful. I want to ensure the feasibility and optimize the architecture before finalizing the system.

    Thanks in advance for your help!

Advertising from our partners