Packaging vision inspection: camera and defect guide

Define defects in measurable terms

Replace words such as “wrong”, “crooked” or “damaged” with dimensions and examples. Specify missing cap, cap height, label edge position, code presence, code text, colour, print registration, seal contamination or component count.

Build a controlled image library from real production variation. Include acceptable extremes so the system does not learn only perfect samples.

Control the image before choosing software

Lighting, lens, exposure, background and product position determine whether the feature is visible. Glossy film, transparent bottles, embossed packs and moving liquids may require dedicated illumination and mechanical stabilisation.

Check depth of field across the full size range and keep the camera-to-feature distance stable.

Set decision and reject rules

Define the region of interest, tolerance and response for each defect. Separate a warning, process trend and mandatory reject. Record why each limit protects the finished-pack requirement.

The reject system must track the inspected item and handle consecutive failures, full bins and downstream stops.

Validate with production challenge sets

Run good and defective samples in random order at minimum, normal and maximum speed. Measure false rejects and false accepts, not only demonstration accuracy.

Repeat after format change, software update, camera movement, lighting replacement or major maintenance. Keep versioned recipes and access control.

Information to prepare

• Defect list with dimensional acceptance limits
• Representative good and bad sample library
• Product position and lighting method
• Camera resolution, lens and field of view
• False-accept and false-reject criteria
• Tracking, reject and fail-safe sequence
• Recipe, user-access and validation records