Practical notes for production, quality and engineering teams: why a check fails, how to choose between methods, and what a camera can and cannot see. Written by Frank Guo.
Insights are short, practical articles on machine vision inspection for South African production lines, written by Frank Guo. Each one answers a question we hear from lines and searches, gives the key points first, compares the options in a table where there is a choice to make, and links to the application, line solution or guide that goes deeper. No sign-up is needed.
HIKROBOT alternatives include Cognex DataMan and In-Sight, KEYENCE sensors and vision systems, and Basler cameras with specified inspection software. The useful comparison is the complete solution for the same task: what it can inspect, who builds and maintains it, and what the quotation includes. HIKROBOT can suit a new integrated station; an established alternative may suit a plant’s existing software, spares and maintenance skills.
A barcode that beeps on the factory scanner can still be refused at a distribution centre. Either its print quality is too close to the limit for the DC’s scanners, or it scans and carries a number the DC does not expect, or a required label is missing. Each has a different fix, and only one of them is about print quality.
Best-before and use-by dates have different purposes and should not be treated as interchangeable label text. For a South African packing line, the practical job is to print the approved date wording and correct value for the product, then verify both against controlled run data. A camera can check the printed code; it cannot determine shelf life or decide whether food is safe to eat.
Packaging changeover verification checks that the approved product, packaging version and variable print data remain matched when a line starts, changes product, reloads packaging or admits rework. The expected identity must come from authorised production data. Test the system with deliberately mismatched packs and confirm that they are held or rejected. This extends the general label-verification SOP to multiple packaging levels; it does not approve artwork, verify ingredients or replace the site’s food-safety programme.
A camera measuring station is accepted the same way as any other gauge: by showing how much it varies on your own parts, compared with your tolerance. The study looks familiar, and an in-line study must include the loading, fixtures, environmental conditions and operator actions that actually vary.
Hikvision’s 2025 Annual Report names Hangzhou Hikrobot Co., Ltd. as a subsidiary and lists HikRobot among its innovative businesses, alongside EZVIZ and HikMicro. HikRobot works in machine vision, mobile robots and articulated robots. Hikvision is a familiar name in South Africa through security cameras; for a factory inspection project, what matters is the industrial camera, optics, software and controls connection, not the parent brand.
HIKROBOT vs Basler is mostly a camera-and-software comparison. Basler documents a broad camera range with the pylon software suite and licensable vTools; HIKROBOT documents cameras with the MVS client, the VisionMaster inspection platform, and its own smart cameras and code readers. A camera captures an image; the station also needs optics, lighting, inspection logic and a controls interface, so compare the whole chain.
For HIKROBOT vs Cognex, first decide whether you need a barcode reader or an inspection system, then compare like with like: HIKROBOT ID readers against Cognex DataMan, HIKROBOT smart cameras or VisionMaster against Cognex In-Sight. This guide lines up the product families as each vendor describes them, then gives a test plan and scorecard so the decision rests on your samples, not on brochures.
Compare HIKROBOT and KEYENCE at the same level of inspection complexity. Both sell AI vision sensors for one simple check, smart cameras for several checks in one view, larger vision systems and fixed-mount code readers. This guide lines up the series as each vendor describes them, helps you pick the device class first, and gives a test plan so the brand choice rests on your samples.
Wrong labels can enter a run at changeover or between spot checks. A label verification procedure should cover line clearance, first-off approval and checks during the run. An in-line camera can check units between manual checks, provided the recipe, image and reject system have been validated.
Oil, chips and inconsistent seating can change the image of a good machined part or hide a real defect. Diagnose the physical cause before relaxing the inspection threshold. Compare clean and production-condition samples, repeat the loading operation, test the lighting and separate a dirty view from an actual defect. Revalidate both acceptable and faulty parts after any change. A lower reject count is not an improvement if the same change also lets unacceptable parts through.
PPECB says its export product inspections include marking requirements as well as quality and phytosanitary aspects. A packhouse should therefore build its carton-check recipe from the applicable commodity and destination requirements and the approved label. Where PUC, packhouse and other traceability codes are required, verify their values against the current run; do not assume one generic field list applies to every export carton.
For a tyre carrying the modern four-digit manufacturing-date stamp, the first two digits indicate the production week and the last two the year. Continental gives 4521 as week 45 of 2021. These moulded characters are not a barcode: reading them automatically needs an image and character-recognition method suited to the sidewall. A tyre barcode-reading station performs a separate traceability task.
A vision station on a sachet or VFFS line can check visible date and batch codes, print registration and visible faults in the seal region. It needs a stable view of the relevant film or finished pack and a way to track rejected units. A surface image does not by itself prove that a seal is leak-tight or that the sachet contains the correct weight.
A short shot is an incompletely filled moulding: part of the intended cap geometry is missing. A camera can detect visible missing material when the feature is exposed, lit and large enough in the image. It cannot infer sealing performance, material strength or the root cause from the image alone. Start by defining the rejectable feature and collecting actual defective and acceptable closures.
Choose the inspection method from the drawing characteristic, the required uncertainty and the way the part can be presented. Vision is useful for repeated measurements of accessible outlines and features. A CMM can address three-dimensional geometry with a suitable measurement strategy, while a functional gauge answers a defined fit question. These methods can work together. Demonstrating repeatability alone is insufficient: reference agreement, loading variation and the acceptance decision also matter.
Visual seal inspection looks for observable faults such as trapped product, folds or misplaced film. Leak testing asks whether a package allows leakage under the selected method, while seal-strength testing measures a mechanical property of the seal. Checkweighing addresses mass. A pack may pass one check and fail another. Choose the combination from the packaging specification and product risk, then validate each method separately instead of treating a camera pass as proof of package integrity.
Most wrong expiry dates are not printing faults. The coder prints exactly what it holds, and nobody changed what it holds at the day or shift change, or the wrong recipe was loaded. A camera stops those packs only if it knows the date that should be printed. This article covers how dates go wrong, where the expected date must come from, how to prove the station works, and what it cannot prove.
Three methods are sold as a fill level inspection system: a camera that looks at the liquid line, X-ray that measures how strongly the contents absorb radiation, and a high-frequency field that the liquid changes. The container decides which one can work. This article explains what each method can and cannot see, how foam and sloshing affect it, where it sits on the line, and why weighing still proves the contents.