A barcode scanner beeps all day — and reads nothing but barcodes.

Ask it for the printed serial number on a label, a batch code, an expiry date or a document reference, and it has nothing to say. Those are printed characters, not encoded data, so there is no symbology to decode. Until now, that meant either typing the value by hand or keeping a separate OCR workflow — slow, error-prone and rarely worth the trouble on a busy line.

Android-OCR turns an Android PDA into an OCR terminal. It runs on the device as a background service, reads printed text through the built-in camera or scan engine, and types the result straight into whatever field has focus. Press the scan key and the value appears — the same gesture your team already uses for barcodes.

Everything is computed on the device. No network, no server, no image ever leaves the PDA.

What it does

Android-OCR is not an app you open and operate — it is a background service that becomes part of the device:

1
2
3
4
5
6
7

Business app (cursor in a field)
   │  press the scan key / send a broadcast / tap the floating ball
   ▼
HP_OCR_Service (resident foreground service)
   │  capture → crop → on-device OCR → rule extraction
   ▼
The result is typed into the business app's field — as natural as a barcode scan

• On-device recognition — the model is exported to ONNX and runs locally. Images and results never leave the PDA.
• Instant results — the service stays resident and the engine is pre-warmed, so a press returns a result in 1–2 seconds without breaking the operator’s flow.
• Works offline — no network dependency. It works in a warehouse corner or on an isolated production line.

Unlike a barcode plug-in, which decodes 1D/2D symbols, Android-OCR reads the printed text on cards, labels and documents — serial numbers, batch codes, dates, and everything else that has no encoding rule to decode.

Two capture paths: camera or scan engine

Both paths feed the same recognition pipeline. You pick one in settings.

📷 Camera mode

• Framing guide — everything outside the frame is dimmed; only the area inside is recognised.
• Three frame sizes (small / medium / large) — the width stays fixed and the height changes, so you can fit a row or a column and exclude background clutter.
• Digital zoom (1.0×–5.0×) and an optional flash, for close-up detail.

🔦 Scan-engine mode

The scan engine frames the shot instead of the camera:

• Leaves the camera free — ideal when the camera is occupied by another app, or absent altogether.
• LED illumination — the scan head’s own LED improves reliability in low light.
• Physical scan key to capture — inside the window, the scan key acts as the shutter.

Three ways to trigger it

The broadcast channel is what makes integration easy — a business app does not need to bundle an SDK. One line is enough:

1
2

Intent intent = new Intent("ocr.scan");
sendBroadcast(intent);   // the PDA opens the capture window in the current mode

Getting the result into your system

Haptic feedback is configurable too (a short buzz on success, two long buzzes on failure) — useful on a noisy production floor.

One app, five screens

Everything the workflow needs sits in a single app:

Recognition engine

The model runs with ONNX Runtime, with OpenCV handling image processing:

• Detect + recognise — DB text detection followed by CTC recognition, with in-house pre- and post-processing.
• Region-only recognition — roughly 37 ms per region; only the framed area is processed.
• Switchable at runtime — move between tiny and small in settings, no reinstall.

Pre-processing and field extraction

A filter chain (white balance, contrast, enhancement — 15 filters in total) can be attached before recognition to handle dark, low-contrast, colour-cast or shadowed captures. A live preview with a manual OCR comparison makes it easy to tune on site.

A regex rule library then pulls only the fields you care about out of the recognised text:

1
2
3
4

Rule:     Serial number
Regex:    [A-Z0-9]{12,18}
Text:     "SN 4SC4-316-abc-123 batch B2026"
Result:   "4SC4-316-abc-123"

Supported devices

Android-OCR currently supports Brady IPS Android PDAs. Tested models:

Using a non-Brady Android PDA, or a model that isn’t listed? We support additional devices through custom development — contact us with your model and use case.

Built for what comes next

Android-OCR runs a standard ONNX model on a standard Android inference stack. That choice matters:

• ✅ Runs on today’s hardware — no special silicon required; mainstream PDAs are enough.
• 🔮 Gets faster on tomorrow’s hardware — the next generation of Android PDAs ships dedicated NPUs. Because the engine uses standard execution providers (NNAPI / ONNX Runtime), those devices accelerate it with no code changes.
• 📈 The model can grow — upgrade or swap the model without touching the hardware you already own. Recognition capability is no longer fixed at purchase.

Pricing

All licences are one-time — no subscriptions, nothing to renew. Additional language versions are $100 each. Full details on the Android-OCR pricing page.

Need a specific output format, a non-Brady device, or another language? Custom pricing starts at $200, depending on the requirement. Tell us what you need.

Requirements and limits

• Platform: Android 9+ (minSdk 29); phone-form PDAs and tablets.
• Scan-engine mode requires a compatible Brady IPS scan engine. Camera mode needs a camera.
• Licence: bound to the device serial number.
• Language: the interface is Chinese today; the English build arrives in October 2026.
• Key injection: on Android, key injection is system-restricted. Accessibility-based injection works today on any Android PDA (including CJK); native keyboard injection is in progress and will improve further.
• Fully offline: no backend, no database, no data leaving the device.

💬 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.