Start With the Container: Three Architectures
The first decision isn't about camera resolution. It's about where labels and codes can end up on your container, and which cameras can see them. Most applications fall into one of three patterns:
| Architecture | Typical container | What it sees | Watch out for |
|---|---|---|---|
| IML inspection (post-moulding) | In-mould labelled cups, tubs and containers | Label position and integrity within the moulded wall, plus moulding defects | Must sit right after the moulding machine. Applied-label systems don't catch IML-specific defects |
| 2-side (front + back) | Flat or oval bottles with front and back labels | Both label panels at speed: presence, position, skew, wrong label, damage | The sides of the container aren't covered. Unsuitable when a round bottle can rotate |
| 360° (multi-camera, stitched) | Round bottles and jars, wrap-around labels | The full circumference, so codes are found wherever they end up | More cameras and lighting, more setup per SKU. Needs good container handling |
Optomech builds all three: the IML Label Inspector, LIS 2S and LIS 360°. For in-mould labels in particular, see machine vision for IML containers.
The Defect List Your System Should Cover
Agree this list with the vendor in writing. It belongs in your URS.
| Defect | Why it matters | How it is usually detected |
|---|---|---|
| Missing / unlabelled | Unidentified product | Presence check in the label area |
| Wrong label (wrong SKU or strength) | Mix-up: the most serious label failure | Artwork pattern match and/or barcode or product-code read |
| Upside-down / double label | Misleading or hidden information | Orientation pattern; edge and thickness cues |
| Shift / skew / tilt | Hidden text, poor appearance, failed downstream reads | Edge position against tolerance bands |
| Flag, fold, crease, half-cut | Damaged information; label may detach | Edge continuity and surface analysis |
| Barcode unreadable or wrong | Dispensing and scanning errors; wrong product data | Decode and compare with the expected value |
| Batch / MFD / expiry print missing, smudged or wrong | Untraceable or wrongly dated product | OCR/OCV against the expected string |
Label Mix-Up Controls: What the Regulations Expect
For pharma lines, label inspection is one layer in a set of mix-up controls. Here is how the relevant texts frame it:
- Revised Schedule M 14.22: roll-feed labels should be used wherever possible, and cut labels kept in separate closed containers.
- Revised Schedule M 18.5.6: online verification of all labels by automated electronic means can help prevent mix-ups. Electronic code readers, label counters and similar devices must be checked to make sure they work correctly.
- Revised Schedule M 18.5.10: significant reconciliation discrepancies are investigated before release.
- 21 CFR 211.122(g): where cut labelling is used, US sites apply one of the listed special controls, including 100% examination by suitable electronic or electromechanical equipment.
- 21 CFR 211.125(c): label reconciliation may be waived for cut or roll labelling when that 100% examination is performed. This is a US provision, so don't assume it applies in India.
Layered controls work best: dedicated roll stock per SKU, line clearance that removes previous labels, a recipe tied to the batch, 100% label verification, proven rejects, and reconciliation backed by inspection counts. Our Schedule M packaging checklist goes through each control in turn.
The Challenge Set to Bring to FAT
Detection claims mean nothing until they are tested against defects you choose. Build a labelled, stored challenge set for each container family:
- A label from the closest-looking neighbouring SKU (same artwork family, different strength or pack size)
- An unlabelled container and a container with a double label
- Labels shifted and skewed to just outside your tolerance, and just inside it (to test false rejects)
- Flagged, folded and half-cut labels
- A damaged or low-contrast barcode, and a barcode with the wrong data
- Batch or expiry print that is missing, partial or smudged, plus a wrong but perfectly legible string
- A good-product run of agreed length at maximum speed, to measure false rejects
Run the set at minimum and maximum rated speed, and confirm that every challenge unit ends up in the reject bin. Keep the set controlled after FAT, because you will use it for routine challenge tests.
Software and Data Features That Matter Day to Day
- Product recipes by SKU, selectable only by authorised users and logged when selected.
- Recent reject images on screen so operators can see why units are rejected.
- Per-batch reports of inspected, accepted and rejected counts, to attach to the batch packaging record.
- Access levels and an audit trail for parameter changes. These are expected for GMP computerised systems.
- Remote support, run under your IT and change-control procedures.
Optomech LIS systems provide a live display of the last 100 defects with images, product-wise recipes, scheduled Excel reports, 3-level access control and pneumatic rejection, as published on the LIS product page. We confirm the data-integrity scope for each order.
Sizing Questions to Answer Before You Ask for a Quote
- Which containers (shape, material, colour) and how many SKUs?
- Which label types: paper, film, transparent, metallic or shrink sleeve?
- Where is variable data printed, by which coder, and where does the expected string come from?
- What is the maximum line speed, and how is container handling done at the inspection point (pitch, guiding, rotation)?
- Where can the station sit: after the labeller, after the coder, before the cartoner?
- What reject method, and how much space is there for a lockable reject bin?
Optomech starting-from price bands for vision inspection systems are on our vision inspection price guide for India. If you are comparing suppliers, use our 15 RFQ questions.