Tracking tea tins into presentation boxes
Tracking tea tins into presentation boxes requires a link between the observed tin, the robot task and the destination pocket. One target ID and one successful placement record.
Labels that must match the product, codes on flexible packs and print on heavy bags. Three stations for poultry and meat packers, sachet and pouch lines and staple food mills.
Meat, staple-food and dry-goods inspection separates product identity, variable print and visible pack condition. Cameras can verify weight-price label data, inspect accessible food-tray seal regions, check codes and visible seals on flexible pouches, and verify printed bag identity and closure. These decisions do not replace weighing, leakage or food-safety tests. Each scope below explains what is visible, how exceptions are isolated and whether a sourced budget or sample-based quotation applies.
Poultry lines run at large volumes: SAPA reports 1 418 766 tonnes of broiler meat (dressed weight) produced in 2024 (SAPA, 2024). Maize meal is the main staple, with bread second (NAMC), and it is packed in printed bags from small to large sizes. Sauces, drink powders, snacks and spices fill sachets, pouches and flow wraps on high-speed packers.
The faults differ by line but share a pattern: a label meant for another product, a code that did not print or smudged on a flexible pack, a bag that left the closer open or with the wrong grade printed. The environments are hard on cameras: cold, wet cutting rooms on meat lines, and flour or maize dust on milling lines. Weighing stays with the checkweigher or the labeller’s scale; the camera checks what is printed and visible.
A meat tray can be glossy and wet; a flour bag can be dusty and uneven; a spice sachet can be reflective and strongly patterned. These conditions change contrast and geometry in different ways. Define the substrate, print method, viewing face and permitted contamination for each recipe. A camera chosen from a clear laboratory label alone has not addressed the production problem.
| Line position | Check | Failure example | Disposition |
|---|---|---|---|
| Meat tray sealing | Tray identity, visible seal-band obstruction and label match | Film wrinkle or product across the seal region | Hold visual fault; validate integrity with the chosen packaging test |
| Maize meal or flour bag closing | Closure appearance, batch/date and bag artwork | Dust hides print or wrong bag stock is fitted | Control dust and compare the approved bag variant |
| Sachet forming and cutting | Code placement, cut registration and visible seal band | A readable code lies partly outside the finished pack | Track each lane and inspect the finished package geometry |
| Case and pallet assembly | SKU, batch and aggregate identity | Different batches merged under one incorrect case label | Hold and reconcile the affected aggregation |
Consider a hypothetical dry-goods line where the code is sharp before the bag is closed but folds into the seam afterwards. An upstream OCR pass does not establish final consumer visibility. Move the check to a position where the finished code can be seen, change the print location or add the necessary downstream verification. For a multi-lane sachet machine, challenge an individual lane fault and confirm that the reject tracking identifies the corresponding finished sachet.
For wet meat packs, compare approved appearance references with real contamination and wrinkle samples in the seal band. A fold can look dramatic yet be outside the functional seal, while a narrow channel can be visually subtle. Quality must define the region and the test method used to decide integrity. Keep checkweighing, metal detection or other hazards controls separate where the product’s risk assessment requires them; a visual pass cannot substitute for those outputs.
Maintain a release record that names the product, recipe revision, packaging revision and the source of expected code data. The camera and coder can agree with each other while both use the wrong order. An independent first-off check against the approved specification closes that common-cause gap. R146 regulations 11 and 12 in the cited dated text address batch identification and date marking; confirm the applicable product rules when approving the artwork. (South African food labelling regulations R146, published 1 March 2010)
Use the food-premises hygiene programme to define access for cleaning, permitted chemicals, window inspection and restart checks. R638 is the local regulatory context; it does not prescribe a universal camera maintenance interval. (South Africa: food premises and transport hygiene regulations R638, 2018) Include an obscured-view challenge after cleaning and a known failure through the reject device. If confirmation fails, hold the affected flow according to the plant procedure rather than counting a software reject as a removed pack.
The commissioning report should preserve separate results for identity, visible defects, text correctness and physical removal. Record acceptable packs incorrectly held as well as faulty packs released, with the relevant tested population for each. Keep the challenge pack identity and image so a future maintenance team can reproduce the test. A model confidence score is useful diagnostic evidence but is not the same as the probability that a customer receives a defective pack.
Estimate the cost of a missed fault at the stage it escapes: filling material and labels added to a defective container, time spent sorting finished cases, destruction or rework of held stock, and the administrative cost of tracing affected batches. An early unnecessary reject has a different cost from an incorrectly released finished case. Use those distinctions to set priorities; no saving percentage is assumed here. The linked production solutions cover station layout and assessment inputs, while this industry guide defines which checks belong at each process stage.
Compare the inspection scope and HIKROBOT configuration. Where a sourced budget is available it is shown in rand excluding VAT; other scopes are quoted after sample assessment.
| Line solution | Faults it catches | HIKROBOT configuration | Indicative budget (excl. VAT) |
|---|---|---|---|
| Weigh price label verification | Missing or wrong labels, wrong product, unreadable barcode, wrong date | Code reader + vision sensor or colour camera, tray pusher | R 160 000 – R 740 000 |
| Pouch inspection | Missing or smeared codes; product in seal, folds, open edges; short cartons | Vision sensor on the code, line scan over seals, air-jet reject | R 80 000 – R 930 000 |
| Maize meal bag inspection | Wrong bag stock for the grade; missing or unreadable code; bag not closed | Vision sensor or 2 cameras in dust housings, bag reject | R 110 000 – R 770 000 |
| Food tray seal inspection | Food reaches the tray flange; Film is folded or off-centre | HIKROBOT area-scan cameras with task-specific lighting and PLC quality release | Quotation after sample assessment |
| Food packaging version verification | The previous product’s labels remain on the line; The barcode is right but the artwork is old | HIKROBOT camera/reader selected for the actual identifying feature with task-specific lighting and PLC quality release | Quotation after sample assessment |
The same inspection applications appear across these lines; each links to the page that explains it in depth.
| Check | Application | What it catches | Where it stops |
|---|---|---|---|
| Label content and match | Label inspection | Wrong label for the product, missing label, unreadable barcode, wrong date | Weight and price accuracy: that is the scale’s job |
| Codes on flexible packs | OCR/OCV and code inspection | Missing, smudged or wrong date and batch codes on film and bags | Grading print quality to ISO standards |
| Visible seal area | Fill level, cap and seal | Product trapped in the seal, folds, open edges | Leaks and seal strength |
| Count per carton | Presence and counting | Short cartons of sachets or packs | Packs hidden under others in a closed carton |
| Pallet and case labels | Carton and pallet verification | Case and pallet labels matched to the order | Contents of a closed case |
The rules require the outcome; a camera is one way to check it on every unit.
The label on a meat tray, the flange of that tray and a printed staple-food bag represent different observable information. Define each acceptance condition rather than reporting one general pack pass.
| Decision or event | Check or action | Acceptance boundary |
|---|---|---|
| Rigid food tray | Visible flange contamination, lid position or film folds | Use the applicable integrity/strength tests separately. |
| Weight-price label | Expected product, displayed fields and barcode data | The camera does not weigh the pack. |
| Staple-food bag | Approved grade artwork, closure and code | Visible closure is not a test of all barrier properties. |
| Flexible pouch | Visible seal region and variable print | Opaque film can hide contamination. |
Validate the relevant pack families, normal production variation and each specified defect. Record results by lane or product family where those conditions change the inspection.
A station after the weigh-price labeller reads the label’s barcode and text and compares them with the product the labeller recipe or PLC says is running. A label for another product, a missing label or an unreadable barcode is rejected before the tray is cased.
Usually, when the code is printed in the same place on every pack and the film does not wrinkle at the camera. Multi-lane sachet machines may need one camera across the lanes or one per lane; the sample test on your film decides.
The established packaging scopes give dated indicative installed budgets, from about R 80 000 for a single vision sensor to about R 930 000 for a station with line scan over the seals, excluding VAT. Those figures apply only to the priced scope described on each page. Newly added scopes are quoted after sample assessment; all projects require their own final quotation.
No. The camera checks print, codes and the closure; the checkweigher or filling scale proves the mass. They work side by side on the same line and can share the reject.
With an IP67 housing, a purge or shroud and a cleaning routine, yes in most mills. Dust that settles as a crust on the window will blind any camera, so the mounting point and cleaning are part of the site survey.
Describe the products, the line and the fault you need caught. We test on your samples and reply with a feasibility note and a budget.