A fixed-mount reader has seen more than its share of barcodes pretending to be fine.

A skewed, faded or smudged code scans today and fails tomorrow — and by the time it fails, the line has stopped, the customer has returned the batch, and the compliance paperwork has already been filed. And the characters on the part — the laser-etched serial, the printed lot number, the moulded text — are not a barcode at all, so the reader has nothing to say about them either.

Fixed-Mount OCR V2 combines two capabilities into one line-side solution:

• Fixed-mount OCR — the reader reads characters, not just symbols.
• Barcode grading — every scan also scores the code’s quality.

One scan produces both content and quality. Quality decline is caught while the code still scans.

Once the code is easy to read, the reader tells you what the part is; grading tells you how long the code will keep reading. That combination is what a real line-side gate looks like.

The two problems on the line

Characters the reader can’t read. A fixed-mount reader decodes barcodes quickly and reliably — but much of the information on the line is not in a barcode at all: laser-etched serials on metal, lot numbers and dates on labels, moulded or stamped text on plastic. These have no symbology to decode. The usual answers are either a heavier vision system (frozen at manufacture) or a large model (too heavy to run on site).

Quality the eye can’t see. A successful scan is not a healthy code. Ribbon ageing, printhead wear, moisture and handling all erode quality gradually:

• An A-grade code today can be a barely-readable C next week.
• By the time scans start failing intermittently, the line’s takt time is already affected.
• In medical UDI and automotive IATF environments, a sub-standard code is a compliance problem, not an inconvenience.

A line doesn’t need a device that merely reads. It needs a station that reads it, understands it and holds the line on quality.

Solution overview

This is not “install some software on a PC”. It is an integrated solution built around a Brady IPS fixed-mount reader: the device handles high-speed capture, and the software releases its full potential.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Photoelectric sensor ──IO trigger──▶ Brady IPS fixed-mount reader (HF700/HF71X/HF81X)
      │
      ▼
┌────────────────────────────────────┐
│        HappyLab vision-server       │
│     bounded queue (burst absorb)    │
│     ├─▶ ① OCR channel ── content    │
│     │   template ROI + regex        │
│     └─▶ ② Grading channel ── A~F    │
│       ISO 15416 / 15415 / 29158     │
└──────────────┬──────────────────────┘
      ▼
 Records: barcode + text + grade + image
      ▼
 Line display / focus entry / traceability export

Why “releasing the reader’s potential” matters: a fixed-mount reader’s strength is speed — IO-triggered, stable cadence, multiple images per second. Traditionally its scope ends at “reads barcodes”. This software gives that speed a brain that keeps up:

• Throughput matches the device — FTP capture → in-memory queue → parallel processing; a single frame completes in roughly 90–300 ms, so the software stays ahead of the device cadence. The throughput ceiling is the reader’s, the software is never the bottleneck.
• No wasted capture path — images go straight from the device to a memory queue; nothing hits disk, nothing waits.
• The barcode comes for free — the filename carries it ({barcode}-{timestamp}-OK/NG), so the decoded value becomes a record field with no extra work.
• Capability extends upward — the reader reads the code, OCR reads the characters, grading verifies quality: one fixed-mount device doing the work of three stations.

OCR: reading the characters

A line-grade capture path. Unlike a handheld, a fixed-mount station is fully automatic: a photoelectric sensor triggers the IO, the reader captures, and the image is pushed over FTP — no operator involved. A bounded queue absorbs irregular feed, dropping the oldest frame rather than blocking the device, and 0/90/180/270° rotation is compensated in software so the mounting angle is unconstrained.

Two-channel recognition engine:

Template ROI — the software finds the region itself. Box the target on a sample image (serial, date, prize — as many regions as you need), learn the ORB features and save it as a template. Each run locates the ROI automatically and recognises only what’s inside it: fast and clean. A tilted part is recovered by ORB + homography to within 2–9 px, and a three-level fallback (template → feature → whole-image) keeps the line running when the target moves.

Regex rules — extract only what you want: any field, priority ordering, JSON import/export. Changing a scenario is a configuration step, not a development project.

Barcode grading: telling good from bad

The same scan that reads content also feeds the grading channel — content and quality from one trigger.

Scores run 0–4.0 and map to A–F, where the overall grade is the worst parameter (the weakest-link rule):

The in-house engine grades a frame in under 100 ms, keeping pace with the line. Dual decoders (zxing with a zbar fallback, auto-upscaled) raise the read rate, and the 2D parameters — FPD, GNU, ANU, UEC — are implemented in-house, which is rare in open source. Reports are visual: location box, scan lines and module-grid overlay make the problem obvious at a glance.

Deeply integrated, not a side tool: the grade lands in the same record as the barcode, the OCR text and the original image. A line display shows a breathing status light — green for OK, red for a poor code — visible from five metres. Print-parameter drift shows up as a trend, so you fix the printer before the customer notices.

A code that grades C, D or F still scans — but it is already a warning: the printhead needs attention, the ribbon needs adjusting. Trend monitoring is what stops a quality incident before it happens.

Background mode: results straight into ERP or Excel

After recognition and grading, the result is typed into whatever system you already use — Excel, ERP, an in-house app — via focus injection (keyboard simulation with CJK support, Tab navigation, auto-Enter and optional prefix/suffix). Records are logged automatically and export to Excel; the line display and the background entry mode run side by side without interfering.

1
2

Sensor trigger → reader scans → OCR + grading → data lands in the sheet
              ↑ seamless end to end, as fast as a barcode scan

Application scenarios

Technical specifications (measured)

Supported devices

Fixed-Mount OCR V2 currently supports Brady IPS fixed-mount readers: HF700 · HF711 · HF810 · HF811.

Using another fixed-mount reader? We support additional models through custom development — contact us with the model.

Pricing

Custom options

All licences are one-time — no subscriptions, nothing to renew. Because multi-lane and Android-console builds depend on your hardware and layout, they start with a conversation: tell us about your line and we’ll come back with a plan and a price. Full details on the Fixed-Mount OCR pricing page.

Free 30-day trial

Try the solution with your own reader and your own parts for 30 days, free. Trials are arranged by contacting us today; an online order system for trials and licences is coming soon.

Barcode quality grading (optional add-on) — please read

That said, “reference-grade” is far from “useless”:

• Line-side screening — filter out obviously poor codes before they reach a professional verifier, cutting verification volume dramatically.
• Trend monitoring — daily sampling surfaces decline early, so ribbon changes and parameter tweaks happen before a customer finds the problem.
• Diagnosis — the report says which parameter failed (contrast, quiet zone, smudge), pointing straight at the cause.

Requirements and limits

• Reader: a supported Brady IPS fixed-mount reader (HF700 / HF711 / HF810 / HF811) with photoelectric IO triggering.
• Host: a Windows or Linux industrial PC. No GPU required for the standard build; multi-lane needs 32 GB RAM and an NVIDIA GPU with ≥ 8 GB VRAM.
• Interface language: Chinese today; the English build arrives in October 2026.
• Grading is reference-grade — not a compliance certification (see above).
• Fully offline: no backend, no cloud, nothing leaving the plant.

💬 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.