What Is the Difference Between Motion Detection and Person Detection?
Motion detection triggers on pixels changing in the frame. Person detection runs a classifier on the camera that decides whether the thing that moved is a person, a vehicle or neither, and only then triggers. The practical difference is the volume of alerts: a parking lot camera on motion detection can generate hundreds of events on a windy night, and the same camera on person detection might generate four.
For a commercial building, that is not a convenience question. Alert volume decides whether anyone still looks at the alerts by week three, and whether the footage you need can be found in under a minute or has to be scrubbed through by hand.
How Each One Actually Works
Motion Detection Is Pixel Change, Not Movement
Classic motion detection compares each frame to the last and raises an event when enough pixels in a defined zone change by enough. It has no idea what caused the change. Things that reliably set it off in a commercial setting include headlights sweeping a wall, rain and snow falling through the frame, a flag or banner, steam from a roof vent, shadows moving across a dock through the afternoon, a spider building a web across the lens at night, and the infrared illuminator switching on at dusk.
It is not useless. On an indoor scene with stable lighting and nothing moving that should not be, such as a server closet or a stockroom after hours, pixel-change detection is reliable and cheap on processing. The failure mode is outdoors and anywhere with glazing, weather or traffic.
Person and Vehicle Detection Classify the Object
Modern commercial cameras run object classification on the camera itself, deciding whether a moving object is a person or a vehicle before raising an event. Because the decision happens on the camera, it does not depend on your recorder or your internet connection. The practical gain is that weather, light changes and animals stop producing events, and that the events you do get carry a label you can search.
That labeling is the part businesses underestimate. Searching six weeks of footage for every person who entered the yard after 7pm is a query against the recorded metadata, not an afternoon of scrubbing. Hanwha Vision, which we lead with as a Gold Partner, builds this classification into the cameras we specify most often, and the same capability appears across the other lines we install.
Line Crossing, Loitering and Direction
Once the camera knows what an object is, the rules get more useful than on or off. A person crossing a line in one direction can alert while the other direction stays silent, which is how you watch an exit-only door without alerting on everybody leaving. Loitering rules alert when an object stays in a zone longer than a set time, which suits a dock apron or a rear alcove. Vehicle rules can watch a gate without alerting on staff walking past it.
What This Means for Your System
Monitoring Costs and Credibility
If your cameras feed monitored video, false events are not just noise. Every event that reaches an operator consumes attention and credibility, and a site that produces constant false alerts gets treated accordingly. Cutting events to the ones that involve a person or a vehicle is usually the single biggest improvement available on an existing system, and on many commercial cameras it is a configuration change rather than new hardware.
Recording Mode Interacts With It
A system set to record only on motion will have gaps wherever detection failed, and those gaps are invisible until you need the clip. We generally record continuously where storage allows and use detection for alerting and searching rather than for deciding what gets written. That separation is also why a badly tuned detection zone can quietly cost you footage. See why some cameras stop recording after a few days for what that looks like from the outside.
Tuning Is Part of Commissioning
Detection rules set from a laptop in a van on install day do not survive the seasons. Foliage grows, racking moves, winter brings headlights into frames that were dark in July. On the systems we maintain under a service agreement, reviewing detection zones is part of a scheduled check, because the alternative is a system that everyone has learned to ignore.
Related Questions
Can I add person detection to cameras I already have?
Sometimes. If the cameras have the processing for it, enabling classification is a configuration change and costs nothing but the time to set up and tune the zones. If they do not, the classification can often be done on a newer recorder or a server instead, at some cost in accuracy and with more load on your network. On older cameras, neither is possible and the honest answer is that the fix is the camera. An audit of what you have will tell you which of the three you are in.
Does person detection work at night?
Generally yes, but it depends on the image the classifier is given. A camera with good infrared illumination or white light that produces a clean monochrome image at night classifies people reliably. A camera staring into a bright security light, or one with the illuminator washing out the near field, will produce an image the classifier cannot work with. Night performance is a lighting and lens question first and a software question second, which is why we check the night image on commissioning rather than only the daytime view.