← All project packs

Visual inspection of components with OpenCV

In-line silkscreen verification with rotational template matching: no proprietary vision PLC required.

Runs on: Raspberry Pi + USB camera, Python 3 scripts (Linux PLC).

The application

A manufacturer of electrical components wanted to verify that every part leaves the line with the correct silkscreen, without investing in a proprietary vision system. A Raspberry Pi with a USB camera runs an OpenCV pipeline (background segmentation, CLAHE, adaptive thresholding) and compares against references using template matching across ±15° rotations; a score ≥ 0.85 means the part is correct. The operator manages references and parameters from a touch GUI with 4 tabs.

Architecture

• 1080p USB camera with manual focus locked by software
• Pipeline: background → morphology → cropping → CLAHE → adaptive thresholding
• Template matching with -15°/+15° rotational search
• PySide6 GUI with camera thread (live only on the active tab)
• Persisted references and parameters (processed PNGs + config.json)

Bill of materials

• Raspberry Pi + USB camera: Industrial Shields controller
• 1080p USB camera with manual focus: image capture
• Touch screen: in-line operation
• Diffuse LED lighting: image repeatability

What is in the pack

• gui-pyside6-camera-thread.py
• json-config-references.py
• opencv-silkscreen-pipeline.py
• rotational-template-matching.py
• Bill of materials + README