AI
Synology Licence-Plate Recognition: Cameras, Plate Recognizer and Capture
Readable plates start with the camera view. Compare native and integrated recognition paths, keep licensing separate, and test the real lane in daylight and at night.
A camera can show a clear vehicle while its number plate remains unreadable because of glare, motion or angle. Adding recognition software to that view may simply produce more uncertain results. Begin at the lane before choosing an analytics licence.
Three Recognition Paths to Understand
| Path | Where processing happens | What to verify |
|---|---|---|
| Supported Synology camera analytics | On the compatible camera, such as BC800Z | Camera firmware, region support, capture conditions and recorder device licensing |
| Compatible DVA system | On the supported deep-learning recorder | Exact DVA model, supported tasks, speed conditions and task resources |
| Plate Recognizer integration | In the third-party recognition workflow | Stream access, processing host or service, middleware, software versions and licences |
These are different architectures, not interchangeable checkboxes. A camera providing ordinary RTSP video does not automatically provide usable plate metadata. Keep the original Synology and Plate Recognizer integration distinct from newer native camera features.
What the Plate Recognizer Integration Does
The documented integration uses Plate Recognizer Stream and middleware to pass recognition events into Surveillance Station, where action rules can associate events with bookmarks. Synology lists Surveillance Station 9.1.4 or later for that integration workflow.
Check the current Plate Recognizer licence, deployment method and processing requirements separately from Surveillance Station camera licensing. Plan secure local access and restricted credentials; do not expose a recorder directly to the internet simply to make an example connection work.
A Useful Plate Event Needs More Than a Text Match
Keep capture, recognition and the operator’s decision connected.
01 Prove the capture
Keep misses in the test recordCheck day, night, approach direction and intended vehicle speed. A parked still is a different test.
02 Choose the recording ecosystem
Video support is not AI supportAn ordinary video stream does not guarantee the same native analytics or event integration on another platform.
What Changes with BC800Z and DVA
Consider its native camera analytics with the supported Surveillance Station workflow and required device licence.
A dedicated LPR camera for a compatible UniFi Protect system. Check the intended installation and recording platform.
The Synology BC800Z adds on-camera recognition. Its published LPR guidance is for vehicles up to 20 km/h under the specified conditions. Confirm supported plate regions and placement requirements. Its smoke and LPR tasks are not simultaneous options.
On-camera analytics and recording-device licensing are separate questions. Synology states that BC800Z requires a separate Surveillance Station device licence; do not carry over assumptions from older BC500 or TC500 arrangements. Similarly, DVA capabilities differ: the DVA1622’s supported LPR scenario is stationary vehicles, rather than the same moving-lane requirement as every other model.
Choose the Recording Ecosystem Before the Camera
A new system can also use a dedicated UniFi AI LPR camera with UniFi Protect. That is a separate design from the Synology and Plate Recognizer paths above. Ordinary video compatibility does not establish equivalent native plate events or search on another recorder.
| Planning point | Synology BC800Z path | UniFi AI LPR path |
|---|---|---|
| Native workflow | Supported Surveillance Station camera analytics | Compatible UniFi Protect system |
| Capture planning | Published edge-LPR capture range: 3–10 m; verify supported regions and speed conditions | Installation guidance describes a 12 m capture distance, up to 5 m mounting height and angles below 25°; tighter angles for high-speed capture |
| Night-time proof | Check the actual plate stream, exposure and supported analytic task | Check LPR Night Vision for the plate type; a separate context camera can preserve a broader night view |
| Acceptance record | Correct plate text, time and linked video | Correct plate text, time and linked video |
These planning values come from different manufacturers’ installation guidance, rather than a controlled accuracy comparison. The UniFi camera’s IR distance and optical zoom are not substitutes for a readable plate at the chosen lane. Inspect the intended entry and exit directions before buying lenses or mounting hardware.
Get the Plate Image Right Before Tuning AI
- Make the intended plate large enough and readable in the actual stream used for recognition.
- Control approach angle and speed; an overview camera may be in the wrong position.
- Check headlights, reflections, rain, shadows and the transition into a covered entrance.
- Check focus, exposure and motion at the capture point using the model’s guidance.
- Verify supported local plate formats instead of assuming one country setting works everywhere.
A demonstration still taken while a car is parked is not evidence for a moving nighttime queue. Test realistic vehicles and directions at the intended operating speed. Use a separate overview view where useful, so the context is retained without compromising the dedicated plate framing.
Write an Acceptance Test That Finds Real Problems
Agree a representative set of authorised test passes. Record whether a plate was visible, detected, read correctly and linked to the correct video. Keep misses and false matches in the record. A percentage without the test conditions or denominator does not help compare solutions.
Check the same process at night and after changes to lighting or camera position. Include partially obscured plates and unfamiliar vehicles where relevant. Confirm timestamps, event search, permissions and the operator’s correction process. Report the test conditions alongside any accuracy figure.
Plan Retention and Access as Well as Recognition
Plate text and related video form a useful searchable record, so restrict access and retain it for a defined purpose. Metadata, extra streams and event frequency also affect the wider recording design. Use the AI-aware NVR storage guide and NVR calculator for the video-capacity estimate, then validate actual usage.
If events will influence a barrier or door, have the authorised access-control design cover safety interlocks, manual override and failure behaviour. A text match is not sufficient authority for an improvised automation. A site proof of concept is a sensible next step when capture conditions or integration behaviour remain uncertain.
Frequently Asked Questions
Can any IP camera recognise number plates?
A readable video stream and a compatible recognition system are both required. Basic IP or ONVIF support alone does not establish plate analytics.
Is Plate Recognizer the same as BC800Z on-camera LPR?
No. The integration and native camera analytics use different processing and licensing paths. Select and verify the intended architecture.
Does BC800Z include a recorder device licence?
Synology states that BC800Z requires a separate Surveillance Station device licence. Check the current camera and recorder terms.
Can I recognise fast traffic with the same settings as a car park?
Do not assume so. The capture and processing limits are model-specific; BC800Z guidance describes up to 20 km/h under its stated conditions.
Why does it work by day but miss plates at night?
Headlights, reflections, exposure, motion and focus can make the plate unreadable. Inspect the actual recognition stream at the capture point before changing analytics settings.
Should a recognised plate open a barrier automatically?
Only within a properly authorised access-control design with appropriate safety and failure handling. Recognition text by itself is not an access decision.
