This articles discusses the importance of using proper motion detection and confidence values to ensure that cloud will send email, push and webhook notifications only when it is needed. For detailed explanation on how to configure analytical rules to account for different zones in the scene please review this article:
https://helpdesk.3deye.ai/#knowledge_base/1/locale/en-us/answer/25
If your camera is missing real events, or sending too many false alerts, the cause is almost always a setting — not a broken camera. This guide explains how detection actually works between your camera and the cloud, the most common mistakes people make when setting it up, and how to fix them.
The short version: when cloud analytics is enabled a camera's video feed goes through two layers of filtering before you see an alert — the camera itself decides what counts as "motion," and the cloud platform then decides how confident it needs to be before telling you about it. If either layer is tuned incorrectly, you'll either miss real events or get flooded with false ones.
1. Two Places Where Detection Can Go Wrong
| Layer | Where it lives | What it controls |
|---|---|---|
| Camera-side motion detection | Inside the camera itself | Decides whether something in the picture has changed enough to count as "motion" at all |
| Cloud-side confidence score | In the video platform | Decides how certain the analytics engine must be before it raises an alert or fires a webhook |
Both layers need to be set up correctly — a perfectly tuned cloud confidence score can't fix a camera that isn't detecting motion properly, and vice versa.
2. The Confidence Score: What It Actually Does
Think of the confidence score as a "how sure do you need to be?" dial for the cloud analytics engine.
- Set too high: the system becomes overly cautious and will skip real events — leading to complaints about missing alerts or webhooks.
- Set too low: the system becomes overly eager and will flag things that aren't really events — leading to false alerts.
Here's what this looks like in the platform — an analytical rule with its target objects (Car, Person, Bicycle) and its Confidence value set to 45:
A confidence value of 45 means the analytics engine will raise an alert once it's 45% or more certain an object matches — a relatively loose threshold that favors catching more events, at the cost of occasionally flagging something that turns out not to be a real match. Raising this number tightens the threshold (fewer false positives, higher risk of missed events); lowering it loosens the threshold (fewer missed events, higher risk of false positives).
3. Camera-Side Motion Detection: Theory vs. Reality
Most cameras offer two settings for motion detection: Sensitivity and some kind of timer. Here's what they're supposed to do — and what actually happens on most cameras.
| Setting | In theory | In reality (most brands) |
|---|---|---|
| Sensitivity | Controls how much a pixel's brightness needs to change to count as motion | Often controls both brightness change and how many pixels must change |
| Timers | Controls far apart in time the changes in brightness can be to be considered the same event | On many cameras (especially budget and common Chinese-manufactured models), this setting does nothing at all |
What this means for you: don't rely on the "timers" field to fine-tune detection — on most cameras, the Sensitivity slider alone is doing all the work. If detection isn't behaving the way the menu labels suggest it should, this mismatch is usually why.
A Simple Way to Picture It
Imagine the camera's image as a grid of small zones. Each time the picture updates, the camera compares the brightness of each zone to a moment ago:
The red zones changed brightness between frames. Sensitivity decides how big a brightness change counts — and, on most cameras, also decides how many red zones like this are needed before it calls it "motion."
4. Timers: Alarm On Delay and Alarm Off Delay (Suppression) can be called something else on different models
ONVIF specifies two timer-style settings meant to filter out brief, meaningless flickers (like a blinking light or an insect passing the lens):
| Timer | Intended purpose | Reality |
|---|---|---|
| Alarm On Delay | Requires the change to last a minimum amount of time before triggering an event, filtering out sudden flickers | Many cameras don't actually have this parameter, or it doesn't function correctly when they do |
| Alarm Off Delay / Suppression | Tells the camera to ignore new motion for a set number of seconds after a short event ends | Works on some models, sometimes is the cause of missed real events |
Why Alarm Suppression Can Hide Real Events
This is one of the most common — and most confusing — issues. Here's the sequence that causes it:
| 1. A false motion event occurs (e.g., lighting flicker) |
→ | 2. Event closes after 10 sec due to no motion. Suppression window starts |
→ | 3. A real event happens 2–3 sec. later (e.g., a vehicle enters frame) |
→ |
4. Camera ignores it (still inside the suppression window) |
The camera is technically working and sensitive enough — the real event is simply getting swallowed by the suppression timer left over from an unrelated false trigger.
5. Case Study: UniView Cameras
This exact problem shows up often on newer UniView cameras (firmware from roughly the 2022 and beyond), under Setup → Intelligent/Events → Motion Detection. The relevant field is labeled Suppress Alarm(s), found under Alarm Parameters next to the Sensitivity slider:
- Sensitivity (top right) is set to 70 in this example — this is a good starting range for most scenes.
- Suppress Alarm(s) is the field that causes the missed-event problem described above. The factory default is often set considerably higher than it needs to be — commonly as high as 15 seconds.
- The fix: lower this value. As a rule of thumb, the lowest value available will ensure the most events are detected, since a high value can cause missed events when two triggers happen close together. In practice, setting it to around 5 seconds has tested well — low enough to avoid swallowing real events, while still filtering out momentary flickers.
6. Finding the Right Sensitivity
There's no universal "correct" number — sensitivity behaves differently across brands, models, and even firmware versions. That said, here's a reliable starting point and process:
- Start at 70–80%. Most cameras perform reasonably well in this range.
- Set it to 75% and let it run for a couple of hours.
- Check the video portal to confirm it's generating events as expected — not too many, not too few.
- Adjust from there based on what you observe.
When the Standard Range Doesn't Work
| Symptom | Likely issue | What to try |
|---|---|---|
| Continuous events even when nothing is moving | Camera is over-sensitive | Lower sensitivity, sometimes as low as 25–30% |
| No events generated at all | Camera has an unusually narrow "workable window" (e.g., only responds between 92–94%) | Try going to 90% then carefully nudge sensitivity up in small (1–2%) increments and re-test each time |
Some cameras truly do have a very narrow range where they behave correctly — one percentage point can be the difference between zero events and constant false events. Patience and small test increments matter here.
7. Setting Expectations
Don't expect a newly added camera to work perfectly right away. This is often true — but not always.
- Out-of-the-box performance is inconsistent. Roughly 60–70% of the time, default settings work fine. In the remaining 30–40% of cases, the camera will either generate no events or constant false ones, especially true when starting with a new to you camera model.
- Motion detection may be disabled by default. Some cameras ship from the factory with motion detection turned off entirely — this must be manually enabled, this is rare but does happen, more often on cameras with a lot of advanced/AI features.
- Tuning is mandatory for monitored cameras. If the camera is used for active monitoring or real security purposes, sensitivity and motion masks must be configured — this step cannot be skipped.
- The platform can't fix bad camera-side settings. Cloud analytics can only work with the data the camera sends it — if the camera-side configuration is poor, no cloud setting can compensate.
Quick Setup Checklist
- Confirm motion detection is enabled on the camera (don't assume it is by default).
- Set Sensitivity to 75% as a starting point.
- If the camera is a UniView model, lower the Suppress Alarm(s) value to around 5 seconds.
- If the camera is Hikvision or Dahua disable all smart motion detection features.
- If the camera is Axis add and start motion detector application.
- Let the camera run for a couple of hours, then review the video portal for results.
- If too many false events: lower sensitivity in small steps.
- If too few or no events: raise sensitivity in small steps, watching for a narrow "workable window."
- Ensure the confidence score is set in accordance with the expected alerts. Most of the time default 60% is acceptable but for cameras with challenging scenes this has to be adjusted.
- For monitored/security use cases, confirm masks and sensitivity have been reviewed — don't leave this on factory defaults.
Still getting missed or false events after following these steps? Contact 3dEYE Support — include the camera ID and a screenshot of your current motion detection settings so we can help pinpoint the issue faster.