SGARIEI

Technical Survey to develop Quality Control Stations for Vision Systems.

  • A technical survey for a quality control station with machine vision systems is the exhaustive process of gathering information, analyzing the production environment, and defining requirements to design a solution that automates product inspection.

  • This survey involves identifying what will be inspected, how it will be illuminated, where the equipment will be placed, and how it will communicate with the rest of the plant.

  • The key components are described below:

  • 1. Definition of the Quality Objective.
  • • Defect identification: What characteristics are being looked for (scratches, dents, missing components, labeling errors).
  • • Definition of "Good" vs. "Bad": Establish precise tolerances. Physical samples of defective and acceptable parts are usually requested.
  • • Inspection type: Dimensional measurement, presence/absence, optical character recognition (OCR), or color inspection.
  • 2. Analysis of the Piece (Object of Study).
  • • Physical characteristics: Material (reflective metal, plastic, fabric), size, color and texture.
  • • Variability: Analyze whether the product changes shape, color, or if there are different models in the same line.
  • • Orientation: How the piece arrives at the station (always the same, random, rotated).
  • 3. Production Environment and Line.
  • • Line speed: How many pieces per minute should be inspected (cycle time).
  • • Available physical space: Measurement of the area where the cameras, lighting and supports will be installed (limited space, vibrations).
  • • Environmental conditions: Presence of dust, humidity, high temperature, ambient lighting (sunlight, ceiling lights) that may interfere.
  • • Rejection system: How the defective product will be removed (pneumatic arm, diverter, line stop).
  • 4. Machine Vision Requirements (Hardware).
  • • Lighting (Critical): Selection of technical lighting (backlight, ring light, dome type) to highlight the defect and eliminate unwanted reflections.
  • • Optics and Camera: Defining the camera resolution (pixel resolution needed for measurement) and selecting the appropriate lens (focal length).
  • • Positioning: Working distance of the camera with respect to the part.
  • 5. Integration and Automation.
  • • Communication: Definition of the communication protocol with the PLC (ProfiNet, EtherNet/IP, Modbus).
  • • Outputs/Inputs: Part presence sensors (trigger) and control signals for the rejection actuator.
  • • User Interface (HMI): Define what information the operator needs to see on screen.

  • Survey Result: A detailed document with the technical specifications that allows the selection of hardware (cameras, lights, lenses) and the development of the appropriate inspection software for the machine vision project.


  • DATA:

  • To develop a machine vision system for quality control, it is essential to collect data from both the production process and specific images of the products. The quality of the data determines the accuracy of the system.

  • Here are the key data points divided by categories:

  • 1. Product Data and Defects (Visual Database).
  • • Quality Images (Good): Hundreds or thousands of images of flawless products to train the model on what constitutes a "correct part".
  • • Images of Defects (Bad): Labeled images of all possible types of defects (scratches, cracks, discoloration, missing pieces, deformations, dirt).
  • • Normal Variability: Images with slight variations that should not be considered defects (changes in ambient lighting, slightly different position, natural textures of the material).
  • • Classification and Labeling: Data must be categorized. For Deep Learning, between 500 and 2000 images are needed, labeled by default category for high accuracy.
  • 2. Technical Data of the Workstation.
  • • Definition of "Defect": Exact technical specifications, critical measurements, dimensional tolerances (maximum size/depth of a crack).
  • • Field of View (FOV): The size of the area that the camera needs to capture to see the entire piece or area of ​​interest.
  • • Required Resolution: Pixels needed to detect the smallest defect (e.g., a 0.1 mm scratch).
  • • Line Speed ​​(Cycle Time): How many pieces per minute/second must be inspected to determine the required processing speed.
  • 3. Datos del Entorno Físico
  • • Lighting: Data on existing and required light. Controlled lighting (diffused, direct, stroboscopic) is usually required to highlight defects.
  • • Positioning: How the part will be presented (conveyor, fixed position, robot) and whether the part moves or is static during image capture.
  • • Connectivity: Interfaces required to communicate the vision system with the PLC (drive, PLC, database).
  • 4. Operational Data and Scenarios.
  • • Actions in Case of Failure: What should the system do if it finds a defect? ​​(activate an alarm, separate the part with a pneumatic arm).
  • • False Positives/Negatives: Define the tolerance level for products accepted erroneously vs. good products rejected (false rejections).
  • Implementation summary.
  • 1. Define objectives: What is inspected (presence, dimension, surface).
  • 2. Image Acquisition: Capture real images of the environment.
  • 3. Labeling: Marking defects in images.
  • 4. Training and Validation: Train the AI ​​model with the data.

  • Additional Requirements:
  • List of processes.
  • List of procedures.
  • List of machinery.
  • List of machines.
  • List of systems.
  • List of departments.
  • List of personnel.
  • Customer list.
  • List of roles.
  • List of production lines.
  • List of production plans.
  • List of main faults.
  • List of current problems.
  • List of losses.
  • List of kpi's main departments.
Money transfers.
Go to top