A barcode that scans is not a barcode that passes.

Skewed, faded or smudged codes scan today and fail tomorrow — and the failure shows up as a stopped line, a returned shipment or a compliance finding, all traced back to a code that looked fine. The problem is that “it scans” is a probability, not a guarantee. Ribbon ageing, printhead wear, moisture and handling all erode quality gradually, and the erosion is invisible to a spot check by eye.

The professional answer is not “scan it and see”. It is to grade the code against the ISO standard and find the problem while it can still be read.

HappyLab’s barcode quality grading engine does exactly that — in software. Upload a label, get a 1D or 2D report with a per-parameter breakdown and an overall A–F grade in seconds.

Try the free online grader →

What the standards measure

ISO defines two core methods for print quality:

Scores run from 0 to 4.0 and map to six grades:

The overall grade is the worst single parameter — the weakest-link rule. One parameter out of spec (an occupied quiet zone, weak contrast) puts the whole code at risk. That is why a code can scan perfectly today and be graded D.

For a 1D code, the nine parameters are: decode, edge determination, symbol contrast, minimum reflectance, minimum edge contrast, modulation, quiet zone, decodability and defects. None of this is guesswork — it is a quantified, repeatable measurement.

What HappyLab built

Commercial verifiers from Cognex, Keyence and Axicon are integrated units: industrial camera, standardised lighting, calibrated optics and dedicated software, typically costing tens of thousands. HappyLab took a different route — implementing the grading algorithms in software, so any computer can run them.

1
2
3
4
5
6
7
8
9
10

Label image in
      │
      ▼
In-house C++ grading engine
 ├── decoding: zxing-cpp primary + zbar fallback (dual engine)
 ├── 1D: ISO 15416, nine parameters
 └── 2D: ISO 15415, eight parameters (module-grid rebuild)
      │
      ▼
Out: symbology / content / per-parameter scores / overall grade (A–F) / annotated image

What makes the engine notable:

• C++ core — a single image grades in under 100 ms.
• Dual decoders — when zxing fails, zbar retries on an auto-upscaled image, raising the read rate.
• In-house 2D parameters — FPD (fixed-pattern damage), GNU (grid non-uniformity), ANU (axial non-uniformity) and UEC (unused error correction) are implemented in-house. These are rarely available even in open source; the common open-source grader covers only a fraction of them.
• Wide symbology coverage — 1D: Code 128 / 39 / 93, EAN/UPC, ITF. 2D: QR, DataMatrix, Aztec.
• Visual reports — location box, scan lines and module-grid overlay make the failing parameter obvious.

Where it pays for itself

Printer-side quality control

Print several hundred labels and a visual spot check will not reveal ribbon ageing or printhead wear. Put a laptop or phone beside the printer, photograph one or two labels per batch and grade them: under a minute per check, and a trend line that shows quality declining before a customer finds it.

Inline with a fixed-mount reader

A fixed-mount reader can hand every captured image to the grading engine. A code that scans but grades C, D or F is a warning: check the print parameters. Aggregate grades per station and you can locate the source of the drift, not just the symptom.

Incoming inspection

Supplier labels vary in quality, and warehouse sampling by eye is not a standard. Photograph, grade and record — one consistent criterion, with an archived report to point to if there is a dispute.

Pre-compliance screening (medical UDI, automotive IATF)

Where quality is mandatory, sending every sample to a professional verifier is slow and costly. Screen internally first and send only what clears your threshold — turning a pass rate based on luck into something closer to a certainty.

Honest positioning: reference-grade, not certification

This is the part we most want to be clear about.

Why commercial verifiers cost what they do. ISO-compliant grading depends not only on the algorithm but on standardised hardware: 45° ring illumination at 660 nm with ±10% uniformity and ambient-light shielding, a standard aperture (0.25 mm by default, chosen by density), periodic calibration against a reflectance reference, a camera of ≥5 megapixels shooting perpendicular within 15°, and a formal three-part report format (grade/aperture/wavelength, e.g. 3.0/06/660). The cost is in meeting those specifications and carrying the certification — not in the software.

What we lack is the standardised light source, aperture and calibration target — those are hardware, not algorithms. But “reference-grade” is a long way from “not useful”:

• Screening — filter out obviously poor codes before they reach a professional verifier, cutting verification volume substantially.
• Trend monitoring — sample daily and catch decline early, so ribbon changes and parameter tweaks happen on your schedule.
• Diagnosis — the report names the failing parameter: contrast, quiet zone, smudge. It points at the cause, not just the symptom.
• Anywhere — incoming inspection or a supplier visit; a phone photo is enough.

We are not trying to replace a professional verifier. We are making it possible for anyone with a computer to screen first.

Use it standalone, or wire it in

The engine ships as a standalone product, and it is built to be integrated:

• Standalone app — a desktop application for graders with no existing system to plug into.
• API — score images programmatically and return the full report, so your MES, ERP or QC system can call it directly.
• Fully offline — no cloud service and no external calls. It runs on your network, on your hardware. Nothing about your labels leaves your site.
• Custom integration — we adapt the input, output format and workflow to your line.

An offline, API-accessible grader is what makes this usable in plants where sending label images to a cloud service is simply not an option.

Pricing

All licenses are one-time — no subscriptions, nothing to renew. Full details on the barcode grading pricing page.

Need it wired into your MES/ERP, a different report format, or an on-premise deployment? Custom work starts at $200, depending on the requirement. Tell us what you need.

Free 30-day trial

Try the grader on your own labels for 30 days, free — or start with the online tool and upload an image right now, no signup required.

Requirements and limits

• Input: an image of the code. For best results, capture it flat, evenly lit and in focus.
• Platform: Windows or Linux; the standalone app and the API run on standard hardware with no GPU required.
• Certification: reference-grade only — not a substitute for a certified verifier (see above).
• Interface language: Chinese today; the English build ships in October 2026.
• Fully offline: no cloud, no external calls, nothing leaving your network.

💬 Questions or a custom requirement?

This global site is brand new, and our online tools are currently in Chinese. Tell us what you need and we’ll set up a fast, tailored solution for you — an English version is coming in the next few weeks.

📧 sales@happylab.me · 🌐 global.happylab.me

HappyLab builds practical AI and browser-based tools for industrial device workflows.