Home / Machine vision / Pharma

Pharma · Batch · DataMatrix

Machine vision for the pharmaceutical industry

On a pharma line, a badly printed batch number, an unreadable code or a carton carrying another product's label is not just waste: it is a batch on hold. With machine vision we check every unit, and any unit that fails is rejected with its photo and its reason.

What we check

Every unit, not a sample

Checks that can be installed separately or together, depending on where the problem starts on your line: at the coder, at the labeller, at the closure or at the end, in the case.

Coding

Batch and expiry date

Reading (OCR) and verification (OCV) of batch and date against the production order. An alert when the print head starts to degrade, before a character becomes unreadable.

Date and batch code inspection →

SKU

The right label and carton

Every carton, bottle or blister carries the label of the current SKU, present, straight and in position. A carton from another SKU is rejected even if it is correctly printed.

Closure

Cap and seal

Cap present, straight and at the right height, and the seal or tamper-evident closure intact and correctly fitted.

Cap and closure inspection →

Sealing

Blisters and sachets

Welds and heat seals with trapped product, creases or incomplete sealing, depending on the pack; with deep learning when the defect varies from one unit to the next.

Seal inspection →

Grouping

Units per case

The number of units in every case or bundle, so that none leaves incomplete for the warehouse or the distributor.

Counting in cases →

Smart machine vision camera with built-in lighting

Serialisation

The code has to be readable along the whole chain

In the European Union, prescription medicines carry a unique identifier in a GS1 DataMatrix code and an anti-tampering device (Commission Delegated Regulation (EU) 2016/161). The pharmacy reads that code before dispensing: if it cannot be read, the pack does not go out.

Your serialisation system generates and records the numbers. On the line we check what that system does not see: that every printed code can be read, with which quality grade, and that it carries the data of the current product and batch. If print quality starts to drop, you know before a code becomes unreadable.

  • DataMatrix reading on every unit, not on the sample
  • Print quality grade, with its trend
  • Batch and expiry date in the code compared with those printed in plain text
  • Seal or tamper-evident closure checked in the same step

Mix-ups

The fault a sample does not show

A carton from the previous SKU left behind in the labeller, or the wrong roll of labels after a format changeover: these faults are rare, but a single unit is enough to recall a batch.

That is why each format has its own recipe, and the recipe knows which label, which text and which code belong to that SKU. At the format changeover the new recipe is loaded in seconds and, from the very first unit, anything that does not match it is rejected with its photo. Line clearance remains your procedure; the camera is the check that does not tire on the last shift.

Rejecter at the exit of the TIVOX Sentinel PLUS inspection tunnel

How it decides

One recipe per format, one record per reject

HIKROBOT smart cameras with on-camera processing. Each format has its own recipe: which areas to inspect, with which tool and which threshold. If anything falls out of range, the unit is rejected and its image is stored with the exact reason.

  • Photo of every reject, with time, recipe and reason
  • Thresholds visible to the operator; only supervisors adjust them, with user login and a log
  • Supervised bypass mode for maintenance
  • Format changeover by loading another recipe, in seconds
  • Integration with the line through digital I/O or Modbus TCP, without changing the machine
  • Instant alerts by email, Telegram or WhatsApp

Why it matters

The evidence audits ask for

In pharma, every decision on the line has to be provable. With every reject stored with its image, time, recipe and reason, and every threshold change logged, an audit or a complaint is answered with data. We help you comply.

Access

Users and permissions

Operator, maintenance and read-only, each with what they are allowed to do. Nobody changes a threshold without leaving their name.

Recipes

Change history

The limits of each recipe and their change history, on demand, for quality and for the auditor.

Evidence

Images for the whole shelf life

Reject images are kept for the whole shelf life of the product: a complaint a year from now is answered with the photo.

More on each check: date and batch, labels and DataMatrix codes, caps and closures and counting in cases. If your current system rejects good units, the false rejects guide explains where they come from. Before asking for quotes, the guide to choosing a machine vision system goes through the questions that decide the system.

How to get started

From your samples to the line

Before installing anything, we know whether your defect can be detected and with what reliability.

  1. Remote feasibility test. You send us photos, videos or samples of good and bad units, and we tell you what can be checked in-line.
  2. Installation study. Inspection points, cameras, lighting, rejecter and how it integrates with your line.
  3. Commissioning with good and bad units. Each recipe is tuned and tested with your own defects, and the tests are recorded.
  4. Tuning with line data. With the photo of every reject, the thresholds are fine-tuned so that bad units do not get through and good ones are not thrown away.

FAQ

Answers to the most common questions

What is machine vision used for in the pharmaceutical industry?

To check every unit on the line, not a sample: that the batch and expiry date are printed correctly, that the GS1 DataMatrix code can be read and with what quality grade, that the label and carton belong to the current SKU, that the closure and tamper-evident seal are in order, and that every case contains all its units. Every unit that fails is rejected with its photo and its reason, and that record serves as evidence in an audit or a complaint.

Which cameras do you use?

HIKROBOT smart cameras, with processing on the camera itself. Each format has its recipe, and they are integrated with the line through digital I/O or Modbus TCP, without changing the machine.

What can be checked on a pharma line?

Batch and expiry date (reading and verification), reading and quality grading of the GS1 DataMatrix, the right label and carton for the SKU, cap and seal, sealing of blisters and sachets depending on the pack, and the number of units per case.

Does it replace my serialisation system?

No. Your serialisation system generates and records the serial numbers. Machine vision checks on the line that every printed code can be read, with what quality and with the data of the current batch, and rejects the unit that fails.

Is every reject recorded?

Yes. Every rejected unit is stored with its image, time, recipe and reason, and threshold changes are logged with the user who made them. Reject images are kept for the whole shelf life of the product.

How does it prevent SKU mix-ups at a format changeover?

Each format has its own recipe, with the label, text and code that belong to it. When the new recipe is loaded, any unit that does not match it is rejected from the very first one, even if it is correctly printed.

Do I need to change my line?

No. The camera connects to the line through digital I/O or Modbus TCP, and the rejecter acts on the unit that fails.

What if my case doesn't fit?

We run a remote feasibility test at no cost: you send us photos, videos or samples of your good and bad packs (if a visit to your plant is needed, travel is charged separately). What cannot be guaranteed, we tell you up front.

Free remote feasibility test

Does every unit leave with its batch and its code?

Send us samples of your packs. We'll tell you what can be checked in-line and with what expected reliability.