バーコードスキャン技術（ネイティブアプリ用）

The honest answer is that the best SDK depends on your accuracy, scale, and support needs, so it's worth testing before you commit to any vendor.

That said, Scandit's SDK processed 80 billion scans in 2025 across more than 170 million active devices, and it runs in production at Walmart, Instacart, FedEx, and Lufthansa. Accuracy stays above 99% on the symbologies and conditions most apps target, and it reads the barcodes that defeat other scanners, including tiny, torn, curved, or shiny codes and ones captured in low light, at steep angles, through motion blur, or from a distance. It supports iOS, Android, web, React Native, Flutter, Cordova, Capacitor, Xamarin, and .NET across more than 20,000 device models. You can test Scandit free for 30 days, with no sales call.

Most teams get to their first scan with Scandit in under an hour. You can drop in a pre-built component like SparkScan, which already handles the camera, scan feedback, and the awkward edge cases, or build a custom integration on the low-level API when you need full control.

Either way, you get demo apps, runnable code samples, a library of UX best practices, and QA guides alongside the reference documentation.

Scandit Agent Skills take this further by letting your AI coding agent (Claude Code, Cursor, GitHub Copilot, Codex, Gemini, and 40+ others) write the integration autonomouslyfor you, including SDK setup, license activation, UI wiring, performance defaults, and the edge cases to avoid. You install the bundle with a single line: npx skills add https://github.com/scandit/skills.

Getting to a working barcode scanner takes three steps: add the SDK to your project, activate your license key, and start a scan session. If you use a pre-built component like SparkScan, the camera preview and scanning UI come with it,. Most teams working with Scandit reach their first scan in under an hour.

The fastest route is Scandit Agent Skills, where you describe your use case in plain language inside your coding agent and it writes the iOS, Android, or cross-platform integration straight into your codebase, always targeting the current validated APIs.

You'll find quick-start guides and samples for every framework in our documentation.

Yes, there's a free 30-day trial that gives you an instant license key with every barcode, label, and ID feature included, and no sales call required to get started.

Scandit offers pricing built for businesses of all sizes, with simple plans for small teams and advanced for enterprises — so you pay for your needs, not someone else’s. Customers can choose between a flat annual fee and usage-based pricing.

Scandit’s barcode scanning SDK, ID scanning SDK, Scandit Express app, and enterprise retail solutions are priced separately. There are different editions (pricing tiers) for different scanning needs.

All editions and pricing models include support with evaluation, integration and rollout plus regular software updates. Free trial access is available for evaluation, and demo apps are available for testing.

Scandit is a commercial enterprise SDK that has been in continuous development for more than 15 years and is used by over 2,100 customers.

In our Product Engineering tests against open-source ZXing, run on a Samsung Galaxy S9 with Code 128, Scandit scanned a barcode sequence in 4.0 seconds to ZXing's 6.93, held a 0% false-positive rate against ZXing's 5%, and reached a maximum range of 168 cm versus 37 cm. On 11 damaged "tough codes," ZXing read 2 while Scandit read all 11.

Against Google ML Kit, Scandit was roughly 2X faster in a proof-of-delivery test and 4X faster on bottom-shelf electronic shelf labels, and it has no per-call barcode limit where ML Kit caps at 10.

Test conditions and SDK versions differ by release, so contact us for current data.

What also sets Scandit apart is that it ships pre-built UI, AR overlays, context-aware barcode scanning, and enterprise support with SLAs and a median first reply of around three hours.

You can read the full write-ups in the ZXing and ML Kit comparison blogs.

Open-source libraries are free and perfectly fine for prototypes, but they leave you to maintain the UI, the edge cases, and the library itself. ZXing is now in maintenance mode and only receives bug fixes, while ML Kit depends on Google Play Services, which rules out many rugged and region-restricted devices.

In side-by-side testing against open-source ZXing, run on a Samsung Galaxy S9 with Code 128, Scandit scanned a barcode sequence in 4.0 seconds to ZXing's 6.93, held a 0% false-positive rate against ZXing's 5%, and reached a maximum range of 168 cm versus 37 cm. On 11 damaged "tough codes," ZXing read 2 while Scandit read all 11.

Against Google ML Kit, Scandit was roughly 2X faster in a proof-of-delivery test and 4X faster on bottom-shelf electronic shelf labels, and it has no per-call barcode limit where ML Kit caps at 10.

Test conditions and SDK versions differ by release, so contact us for current data.

What also sets Scandit apart is that it ships pre-built UI, AR overlays, context-aware barcode scanning, and enterprise support with SLAs and a median first reply of around three hours.

You can read the full write-ups in the ZXing and ML Kit comparison blogs.

Yes. Because Scandit is camera-based scanning software, it turns the device camera into a high-performance scanner across more than 20,000 device models, including rugged Android handhelds such as Zebra TC-series devices as well as consumer phones and tablets. Since it doesn't depend on Google Play Services, it also runs on locked-down and region-restricted hardware.

All image processing happens on-device by default, and scanning works fully offline.

Scandit MatrixScan captures every barcode in view from a single press and adds AR overlays so workers can pick, count, and flag right on the screen. Customers report 20X faster inventory counting, 50% less time spent processing shipments, and a 30% higher pick rate.

There's no theoretical limit on the number of barcodes per scan, and pre-built components and customizable AR overlays for counting, finding, batch capture, and AR handle the workflow and edge cases for you. The scan results then pass straight into your inventory, picking, or audit systems in real time.

Accuracy stays above 99% on the symbologies and conditions most apps target. It reliably reads tiny, torn, curved, and shiny barcodes in low light, at steep angles, and through motion blur, at distances up to 168 cm in testing. And if a barcode is too damaged to decode, it falls back to reading the printed digits with OCR with no input needed from the user.

The AI does specific work here: context-aware scanning locks onto the barcode the user means to scan and ignores the others in frame, which cuts accidental scans to near zero, while behavior analysis tells a deliberate re-scan from an accidental dwell instead of relying on fixed timeouts.

The SDK runs on native iOS and Android, web and JavaScript, React Native, Flutter, Cordova, Capacitor, Xamarin, .NET, Titanium, and Linux, covering more than 20,000 device models. On the barcode side, it reads every major 1D and 2D symbology, including UPC, EAN, Code 128, Code 39, ITF, QR Code, Micro QR, Data Matrix (including direct part marking), PDF417, Aztec, DotCode, MaxiCode, GS1 DataBar and Composite codes, and the main postal codes. All image processing runs on-device by default, and the SDK is ISO 27001 certified.