Short answer
A camera with smart analytics can tell whether movement in the image is a person, a vehicle or a swaying branch. That way, you get a notification on your phone only when something real happens. Rules such as line crossing, intrusion and loitering time are built on this distinction; when installed with the right angle and lighting, it gives far fewer false alerts than conventional motion detection.
Why is conventional motion detection so tiring?
In older cameras, motion detection looks at changes in the image pixels. A tree swaying in the wind, an insect passing the camera at night, a cloud's shadow, raindrops and headlights also change the pixels. The result is dozens of meaningless notifications on your phone all night.
After a while, the user turns off notifications. The camera keeps recording, but no one learns about an event right away anymore. Smart analytics was developed to solve this problem: the camera evaluates not the movement but what is moving.
How does person and vehicle detection work?
Smart cameras recognize the shapes of people and vehicles using software trained on a large number of sample images. This is done on the camera's own processor or in the recorder. When the camera classifies a shape as a person, it tags the event; you can choose to get notifications only for people or only for vehicles.
These tags also help when searching recordings. Instead of watching a whole night's recording from start to finish, you list only the moments when a person passed. Some systems also let you search with extra details such as clothing color or vehicle type. Which devices have these features varies from model to model and should be asked about at the quote stage.
Which rules can be set?
On top of person and vehicle detection, rules are set up with lines and areas you draw on the image. The most commonly used are:
- Line crossing: an alert for a person or vehicle crossing a virtual line drawn on the image in a given direction. Suitable for garden walls, fences and the front of warehouses.
- Intrusion: an alert for a target entering a drawn area. Used for parking lots after closing, fields or construction sites.
- Loitering: an alert for a person who stays in an area longer than a set time. Useful in front of ATMs and at door entrances.
- Entering and exiting an area: used to count the people entering and leaving through a given door.
- Left or missing object: an alert for a bag left in an area or an item taken from its place. Installation and setup are more delicate.
Expert note: Don't keep a rule on around the clock. Getting an alert for every employee passing the warehouse during working hours makes notifications meaningless again. Schedule rules for after closing time.
Smart analytics can be wrong too
If the analytics software does not see the target large and clear enough in the image, it cannot classify it. A small human shape in the distance may not be detected, and a dog may sometimes be taken for a person. At night, when the image turns black and white under infrared light, the software's job gets harder.
Most false alerts and missed events come from the installation. If the camera looks at the target from too high and straight down, the human shape is distorted. If the line is drawn at the farthest edge of the image, the target is too small there. If the lighting is insufficient, the software cannot find detail in the image.
So when choosing a camera for smart analytics, look not at the catalog but at a test in your own scene. After installation, have someone cross the line a few times by day and by night, and check whether the alert comes. Repeat the test on a windy day too; you'll see how the system behaves when branches and shadows move the most.
Tips for a correct installation
Place the rule line neither too close to nor too far from the camera. A medium distance, where the target appears at a clear size in the image, gives the best result. Also position the camera so that it views the target's direction of travel from the side as much as possible; a person walking toward the camera seems to cross the line later.
At night, cameras with white light or strong infrared illumination give more consistent results. If there is moving vegetation in the garden or at the edge of a field, draw the line away from the branches. On Şanlıurfa's dusty days, a dirty lens also disrupts the analytics, so have the lenses of outdoor cameras cleaned at regular intervals.
After setting up the rules, review the first week's notifications. Checking at what time and from which camera meaningless alerts came, then moving the line or lowering the sensitivity, solves most problems. Systems left without this fine-tuning soon become as tiring as old motion detection.
Using it together with alarms and sirens
The most powerful use of smart analytics is linking the alert to another system. When a line crossing rule is triggered, the siren and light on the camera can go off, an input can be sent to the alarm panel, or the garden lighting can turn on. Whether the camera and alarm panel support this integration depends on the brands used.
When the light comes on and a warning sound is heard the moment someone approaches the fence, the incident is often over before anyone gets inside. But the rule must be set correctly so that the siren does not disturb the neighbors for no reason.
Smart analytics does not replace a burglar alarm. The alarm gives a definite signal when a door or window opens, while the camera warns of someone approaching outside. Used together, they give you earlier and more reliable warning.
Who is it a good fit for?
Homes with large gardens, warehouses, construction sites, fields and businesses that are empty after hours benefit most from smart analytics. No one should be in these places at night; every human shape that appears in the image is an event that needs attention.
In a store that is busy all day, however, a person detection alert is meaningless. There, analytics is more useful for searching recordings and counting.
When deciding, ask yourself: if a notification came to my phone at night, would I get up and check? If the answer is yes, that notification needs to be real, and that is what smart analytics is for. If the answer is no, analytics still shortens the time spent searching recordings, but what you really need is cameras with good angles and enough recording time.
In the systems it installs, Desa Bilişim sets the analytics rules after seeing the site and according to your working hours.
Frequently asked
Can I add smart analytics to my existing cameras?
Some recorders run the analytics on their own processor; in that case, person and vehicle detection can be used with existing cameras too. If the recorder does not support this, you need a new recorder or camera that does the analytics. With old analog cameras, image quality may limit the analytics.
Will a smart camera mistake cats and dogs for people?
In a well-installed system this is rare, but it does not disappear completely. If the animal is very close to the camera or the night image is weak, misclassification can happen. Drawing the line a little away from the ground, at a distance where the target is clearly visible, reduces this risk.
Are face recognition and person detection the same thing?
No. Person detection tells you there is a person in the image but does not know who it is. Face recognition compares the person against a stored list, needs different hardware and is subject to stricter rules regarding personal data.
Get only real events on your phone
Let us see your site and set up the analytics rules for your working hours and surroundings.
