To verify ANPR country compatibility before ordering, test the exact proposed camera and recognition configuration against real local plates, then repeat the check in a representative lane. Agree on the plate types, test conditions and scoring rules first. Keep the original images, raw recognition output and failures—not just a supplier’s demonstration or one headline accuracy figure.
This guide is for distributors, system integrators and project buyers evaluating a parking entrance. It provides a compatibility brief, a sample matrix and a passage record you can use with suppliers. It is a proposed purchasing workflow, not a report of a completed SHUNJIE recognition test.
1. Replace “supports our country” with a plate-type checklist
A country name is a starting point. Your site may receive ordinary cars, two-line plates, motorcycles, temporary registrations and vehicles from neighbouring countries. A successful demonstration of one format does not answer whether the proposed configuration can handle the rest.
Ask for a written answer for each plate category: supported in the proposed configuration, needs validation, or not supported. Add the evidence reference and any restriction. The Carmen ANPR engine manual, for example, distinguishes country information from plate types and regional engine settings. This illustrates why a country label alone is insufficient; it does not establish compatibility for a different manufacturer’s product.
| Ask the supplier to identify | What your answer should specify |
|---|---|
| Country and relevant state or region | The actual jurisdictions expected at this site, including cross-border visitors if relevant. |
| Vehicle and registration categories | Ordinary cars, commercial vehicles, motorcycles, temporary or special plates actually expected—not a list of every possible global format. |
| Plate layout and output | Single or multiple lines, character sets, front/rear view, and whether country or region must be returned as a separate field. |
| Operating envelope | Lane direction, stopping or moving traffic, expected speed range, capture position, lighting and known obstructions. |
Separate reading the characters from identifying their jurisdiction. If identical character strings could refer to different vehicles in your project, agree how the access database distinguishes them. Do not assume a country field exists or that the parking application uses it.
2. Freeze the configuration you are buying
A file demonstration on a supplier’s computer and an installed camera can use different recognition software. Record the configuration before comparing results:
- Camera or integrated-terminal model, lens and capture direction.
- Device firmware, recognition engine version and country/region package.
- Required licences and where recognition runs: device, local server or external service.
- Capture area, exposure/illumination settings, confidence filters and event timing.
- Exact or approximate list matching, duplicate-event handling and relevant adaptive settings.
- Export format and the system that receives the recognition event.
Ask the supplier to mark settings that are unavailable or not applicable. Keep a dated configuration export where supported, or an agreed settings record. If the supplied product is an integrated ALPR display terminal, verify that complete unit and its software—not only a separate recognition application.
Changes are normal during commissioning, but they must remain visible. A firmware, engine, regional-package or threshold change creates a new test configuration. Preserve the earlier result and run a documented comparison rather than silently replacing it.
3. Build a real sample matrix, including the awkward cases
Collect images or video you are authorised to use. Keep the original files, resolution and timing. Remove unrelated personal information from what you share, but do not repaint, replace or reconstruct the plate characters being tested. Agree who can access the samples and when they will be deleted.
Use two complementary sets: a consecutive or otherwise documented sample of normal traffic, and an additional set for important uncommon cases. Report them separately so a deliberately difficult challenge set is not presented as the site’s normal traffic mix.
| Sample group | Include when relevant | Keep with the result |
|---|---|---|
| Routine local traffic | The plate formats and vehicle classes expected most often. | Selection method, distinct vehicles, passages and entry/exit direction. |
| Less common formats | Two-line, motorcycle, commercial, temporary, special-category or neighbouring-country plates. | A separate count and result for each required category. |
| Lighting conditions | Day, night, headlights, backlight, shadows and wet reflective plates where these occur. | Time, conditions, capture position and original evidence. |
| Real lane behaviour | Turning approaches, different stopping positions, moving vehicles and nearby traffic. | Lane, direction, observed movement and the camera configuration. |
| Operational exceptions | No visible plate, obscured plates or a format outside the agreed support scope. | The exception and expected assistance path; do not score an invented plate as a successful read. |
Agree the sample size around traffic diversity, project risk and the strength of the decision you need to make. A small set can expose an obvious incompatibility; it cannot establish reliable performance across an entire country. Repeated passes by the same few cars should be identified, not described as hundreds of different vehicles.
AI-generated plates, edited numbers and reconstructed images are unsuitable as acceptance evidence. They can illustrate a process, but they do not establish how the system handles real local plates and real capture conditions.
4. Use file testing to screen compatibility, then trial the lane
Start with an independent file test
Use one group of samples for setup and troubleshooting, then keep a separate evaluation group out of that tuning process. Prepare a human-checked answer sheet independently of the ANPR output. Resolve uncertain labels or mark them as uncertain before scoring.
For the initial evaluation run, retain every input and raw output, including no-result and wrong-result cases. Ask the supplier to identify any manual intervention. If settings are adjusted after failures are examined, keep the first result and label the next run as a retest. A final check on fresh, previously unused samples helps show whether the change generalises beyond the examples used for adjustment.
Record the kind of input actually tested. Success on a tightly cropped, clear plate image is evidence about that input; it does not demonstrate that a camera can capture a moving vehicle at your gate.
Repeat with the proposed equipment in a representative lane
Use the agreed hardware and software configuration, realistic approach and stopping positions, and the important lighting and plate categories. For a low-volume site, plan additional controlled passes to cover missing conditions, recording them separately from routine traffic.
Keep an independent passage log or suitably positioned reference recording so missed vehicles remain visible even when the ANPR event list contains nothing. Synchronise clocks closely enough to associate each passage with its evidence, and document any offset.
The UK Home Office’s ANPR performance guidance describes ground-truth comparison, separate capture/read measures and day/night evaluation. Those methods are useful references. Its law-enforcement thresholds and sample requirements are not universal acceptance rules for commercial parking.
5. Score complete plates with clear denominators
Define an eligible passage and a successful capture before the run. State the required direction, plate visibility and operating envelope. Keep all observed passages and exclusions in the record; do not move a difficult but in-scope failure outside the denominator after seeing the result.
For the comparison below, a capture means the system has retained an identifiable image of the target vehicle’s plate associated with that passage, whether or not the text is right. A neighbouring vehicle’s plate does not count. A correct read means the whole required plate string matches the independently checked answer. Agree any permitted treatment of spaces or punctuation first. Do not silently treat a wrong letter or digit as correct.
| Measure | Numerator / denominator | Question answered |
|---|---|---|
| Capture rate | Eligible passages with a captured plate image / all eligible passages. | Did the system obtain the plate evidence? |
| Complete-read rate after capture | Correct complete reads / eligible passages with a captured plate image. | How often was the captured plate read correctly? |
| End-to-end complete-read rate | Correct complete reads / all eligible passages. | How often did a passage produce the right full plate? |
| Jurisdiction accuracy, if required | Correct country/region outputs / eligible passages requiring that output. | Was the required jurisdiction field also correct? Keep missing outputs in this denominator. |
Illustrative arithmetic, not a measured product result: 500 eligible passages produce 490 captures and 475 correct complete reads. Capture rate is 490/500 = 98%; complete-read rate after capture is 475/490 ≈ 96.9%; end-to-end complete-read rate is 475/500 = 95%. All three describe the same example, but answer different questions.
Report both the overall result and the required plate-category/day/night groups. Include the counts as well as percentages. Keep missed captures, incorrect reads, captured images with no text result, uncertain ground truth and excluded cases identifiable. Do not use video frames or multiple updates from one passing vehicle as separate successful passages.
A confidence value attached to one output is not the measured correct-read rate of your project. Its meaning depends on the vendor, field and software version. Ask what it represents and how filtering changes which events you receive; avoid comparing confidence numbers across different systems as though they were the same measurement.
6. Check authorization separately from recognition
A barrier opening is not proof that every character was read correctly. Access software may use approximate matching, another credential, an operator action or an existing session. Record the raw plate text separately from the permission decision and release command.
In a supervised trial with appropriate site controls, test an authorised vehicle, an unlisted vehicle and an expired permission. To check a one-character mismatch, use an isolated test list containing an entry one character different from a legitimate test vehicle’s plate; ensure the actual plate has no permission and verify the expected rejection. Restore the test configuration afterwards. Also check the agreed response to no-read and duplicate events. Do not modify road plates to create a test or bypass safety devices.
Calculate the false-authorization rate as unauthorised test passages granted permission divided by all unauthorised test passages, stating which cases were exercised. Record authorised passages incorrectly denied separately. Keep both measures separate from OCR accuracy. No errors in a small test does not establish a zero-error guarantee.
Where the equipment provides reliable feedback, compare the release command with actual barrier status and the passage record. Otherwise state that the command was verified and the physical outcome was checked by observation. A relay command alone does not prove that the barrier opened or the vehicle passed.
7. Keep one traceable record per passage
The following is a suggested worksheet structure for the buyer and supplier to complete. Use your own case IDs; do not replace the raw files with a summary spreadsheet.
Case ID / run ID: Sample group: routine-challenge / setup-independent evaluation Lane / direction / time / configuration ID: Day-night / weather / lighting / movement: Eligible-excluded / exclusion reason: Vehicle class / country-region / plate category: Reference plate text / evidence file / label certain-uncertain: Original ANPR image or video reference: Raw returned plate / country-region / event ID: Capture: yes-no / complete read: correct-wrong-no result Expected permission / actual decision / manual intervention: Release command / observed or reported barrier outcome: Failure category / corrective action / retest reference:
When a test fails, separate an unsupported format from poor capture, a trigger issue, a software filter or a downstream integration fault. SHUNJIE’s LPR camera fault diagnosis guide offers a related troubleshooting starting point. The failed case should remain in the test record even after the cause is corrected.
8. Turn the evidence into an order decision
Agree the required categories, test conditions, performance measures, thresholds and exception handling before the purchase decision. The thresholds should reflect the site’s consequences and support arrangements, rather than a percentage copied from a brochure.
| Decision | What the evidence should show |
|---|---|
| Proceed within an agreed scope | The proposed configuration meets the agreed criteria for the required categories and lane conditions; limitations and assistance procedures are accepted. |
| Retest before committing | A needed category or condition is missing, or a proposed correction requires fresh evaluation. Record the owner, configuration change, new samples and decision date. |
| Change the solution or scope | A mandatory plate category or site condition remains unsupported, or failure handling is unacceptable. Consider a different capture arrangement, verified configuration or complementary credential. |
Attach the approved configuration and result summary to the supplier response. Identify untested categories, required licences, outstanding development, additional costs and responsibility for later updates. Agree when a software or hardware change requires regression testing. Pre-order validation reduces uncertainty; it does not replace final commissioning and acceptance at the installed site.
Prepare a compatibility brief for SHUNJIE
For a project discussion, send the destination country and relevant regions, required plate categories, expected vehicles, entry/exit views, lane photographs or a plan, operating conditions, and the exact outputs your access software needs. Include authorised original sample files through an agreed sharing method, with a sample index and your proposed scoring requirements.
Use the enquiry form on the SHUNJIE ANPR camera page to discuss the proposed model, configuration and validation scope. Ask which requirements are already supported, which need testing and which are outside scope, then confirm the sample-test arrangements. Country compatibility and project performance should be confirmed for your configuration and evidence, rather than inferred from a general country list.
Featured image: AI-generated illustrative Great Britain parking scene with generic equipment and a simulated plate. Not an actual SHUNJIE installation or recognition test.
Featured image: AI-generated illustrative Great Britain parking scene with generic equipment and a simulated plate. Not an actual SHUNJIE installation or recognition test.
