AI Person Detection Cameras, Explained in Plain Language

If you've ever owned a basic motion-detection camera, you know the problem: your phone buzzes all day for swaying branches, passing headlights, rain, snow, shadows, and the neighbour's cat. After a week, most people turn notifications off entirely, which defeats the point.

AI person and vehicle detection fixes that. Here's how it works and how to get the most from it.

Motion detection vs AI detection

Traditional motion detection compares video frames and triggers when enough pixels change. It can't tell *what* changed. A tree moving in the wind looks the same to it as a person walking up the driveway.

AI detection runs a trained model, either on the camera itself or on the recorder, that recognizes shapes. It classifies what it sees, typically as:

• Person
• Vehicle (cars, trucks, sometimes bikes)
• On some systems, animals or packages

You then choose which types trigger recordings and alerts.

Why it matters

• Far fewer false alerts, so you actually pay attention to notifications
• Faster review, because recordings are tagged by what was detected
• Smarter search, letting you pull up "all people at the back door last night" in seconds (see searching footage by description)
• Better use of storage, since event recordings are more meaningful

Setting it up well

Detection zones

Draw zones on the camera view where alerts should trigger, such as your driveway or walkway, and exclude the public sidewalk or street. This cuts alerts from passers-by and respects neighbours' privacy.

Line crossing and intrusion

Many systems let you draw a virtual line (for example across a gate) or an area (a backyard). The camera alerts only when a person crosses the line or enters the area, not when they walk past.

Schedules

Set person alerts to be active overnight or while you're away, and quieter during the day when family members are coming and going.

Sensitivity

Start at a medium setting and adjust over a few days.

Camera placement affects AI accuracy

AI needs a clear view to recognize people and vehicles:

• Mount cameras roughly two to three metres high, angled down but not straight down
• Avoid pointing directly into bright lights or the rising or setting sun
• Make sure the area is lit at night or the camera has good night vision
• Keep people at a reasonable size in frame; tiny figures at the edge of a wide view are harder to classify

Our guide on where to place security cameras covers this in detail.

Limits to be aware of

AI detection is very good, but not perfect:

• Heavy snow, rain, or fog can reduce accuracy
• A person partly hidden behind a car or bush may be missed or detected late
• Very distant subjects may not be classified
• Spider webs across the lens can trigger events at night

That's why we still recommend continuous recording to a local NVR alongside AI events. If an alert is missed, the footage is still there. See NVR vs cloud storage.

Privacy considerations

AI detection identifies *types* of objects, like people and vehicles. Some systems offer facial recognition; if you consider it, think carefully about privacy and applicable law, especially for businesses. For most homes, person and vehicle detection is all you need.

How we set it up

With our security camera installation, we configure AI detection zones, line crossing, schedules, and alerts for each camera, and fine-tune them with you so your phone buzzes only when it matters.

Frequently asked questions