AI cameras
Cameras that know what they are looking at
They do not say “motion”: they say person, vehicle or animal, with a confidence level, and where — the core of any smart CCTV installation for a villa. They tell you where. That difference is what eliminates false alarms — and what stops a wild boar waking you at 3 a.m.
The three levels of analytics
Person/vehicle filtering
Removes leaves, shadows and light changes. Today's minimum acceptable level — but on its own, not enough in a Mediterranean garden.
On-camera classification
Person, vehicle and vehicle type, processed on the device itself — no server, no cloud.
The animal category
An “animal” event is recorded but never wakes you. In this province that is not a luxury — it is a specification.
Where it fits in your system
New system
The AI camera is the verification layer: radar or LiDAR detects, the camera confirms it is a person, and only then does the alert fire. The full design logic is on the perimeter security page.
Existing system
An AI recorder or analytics appliance adds classification to your current cameras. It is the heart of every upgrade: maximum gain, minimum disruption.
Frequently asked questions
Which camera tells people from animals?
A camera with classification analytics: it identifies person, vehicle and animal separately and alerts only on what you choose. It is not just any camera with “motion detection”.
Does AI stop garden false alarms?
That is its main job: it discards leaves, shadows, light changes and animals. Combined with the radar's filters, it cuts false alerts to a minimum.
Do I need internet or cloud for AI?
No: the analytics we install run on the camera or the recorder, with no compulsory cloud. It works locally.
Does the AI camera record wild boar?
Yes, but as an “animal” event: it is recorded and documented, without firing an alert at three in the morning.
Can I search recordings by object type?
With the advanced levels, yes: search “people” or “vehicles” in a time range, and even by attribute, depending on the equipment. Useful after an incident.
Does AI work with the cameras I already have?
In many cases, via an AI recorder or an analytics appliance that adds classification to existing ONVIF cameras. It is the most economical entry point.
Is AI better on the camera or on the recorder?
On the camera it gives lower latency and reliable relay output for the alarm; on the recorder it converts cameras without AI at a single point. Chosen by case — they do not compete.
Does it tell vehicle types apart?
The mid and high levels distinguish car, van, motorbike, and do line-crossing, intrusion or loitering. It depends on the equipment.
Does the AI read number plates?
The advanced levels do (LPR), useful at gates and private drives. It is considered where the case justifies it, with data-protection safeguards.
Is it reliable at night?
It works on the image the camera produces: with adequate lighting or infrared, yes. Without light, it is paired with thermal to detect and a visible channel to verify.
Who can see my images?
You, as the owner. We configure the system and do not need access to your recordings to maintain it; remote access is by controlled VPN, not by opening ports.
Is the animal category essential in Málaga?
In this province, yes: urban wild boar are documented (Marbella 40%, Málaga 30%, Fuengirola 22.5%, Benahavís 7.5% of the study's animals). Without an animal category, it wakes you.
How much does an AI camera cost?
It depends on the analytics level and on whether existing equipment is reused. It is sized in the fixed quote after the visit.
Does AI replace radar or LiDAR?
No: AI verifies, radar and LiDAR detect. The combination is what eliminates false alarms and gives reliability.
Does it comply with data-protection law?
It is configured under the LOPDGDD: a sign, angles limited to your property, one-month maximum retention and microphones off by default.
What is the difference between motion detection and AI classification?
A motion detector reacts to any change: shadows, headlights, rain, animals. AI classification identifies what each object is — person, vehicle or animal — and alerts only on what you choose. It is the difference between a system switched off in frustration and one you can trust.
Why does the animal category matter so much here?
Because urban wild boar are a documented reality: a study by the Málaga College of Veterinarians with the University of Córdoba located captured urban boar in Marbella (40%), Málaga city (30%), Fuengirola (22.5%) and Benahavís (7.5%). A boar far exceeds the classic detectors' “pet immunity”; the animal category records it without waking you.
Do I need new cameras to get AI?
Not necessarily: an analytics appliance or AI recorder can add classification to existing ONVIF cameras. If your cameras produce a decent image, that is the economical route.
Does AI work at night?
It works on the image the camera produces: with proper lighting or infrared, yes. Where there is no light at all, the right pairing is thermal for detection plus a visible channel for verification.
Do my images leave the house? Who can see them?
The analytics we install run on the camera or the recorder — no compulsory cloud. You, as the owner, are the data controller; we configure the system and do not need access to your recordings to maintain it.
Data protection
Cameras that obey the law, not dodge it
A camera that sees beyond your fence can become a legal problem instead of protection. Spanish data-protection rules (LOPDGDD, art. 22) are clear: a visible information sign, a frame limited to your property and access points, and image retention of one month at most unless required otherwise.
We treat this as part of the project, not as paperwork afterwards. We minimise the field of view to your plot, place the mandatory signage and set retention to the legal limit. In your area, where many plots border the street or a neighbour, framing well from the start avoids the complaints that year after year rank among the most frequent before the data-protection agency.
Recording well is also recording within the law: an image obtained correctly is usable; one obtained by intruding on a neighbour can turn against you.
Zoning
Zoning: making an alarm say where
A house protected as a single block can only say “something is happening”. A house divided into zones says “someone is on the north line, by the pool”, and that difference is what lets you decide in seconds instead of stepping out blind.
We divide the your area plot into logical zones — perimeter, garden, access points, the immediate surround of the house — and each has its own rule: what is watched, at what hour, at what sensitivity. At night the outer perimeter can be armed while you move around inside; when away, everything. The same installation protects differently depending on the moment.
Zoning also narrows the alert: instead of one sensor tripping on everything, each zone reports separately, and the image tells you exactly where to look.
Deterrence
Deterring beats recording the burglar
The best incident is the one that never happens. An intruder sizing up a house looks for the easy way in, and a plot that visibly detects, lights up and reacts is a plot that gets skipped for the next one. Visible deterrence does more work than any recording after the fact.
In your area we pair early detection with zoned response: when something enters the perimeter, that zone's lighting comes on and, where appropriate, the siren marks presence. It is not theatre: behind it is a real detection that has already classified this as a person, so the reaction does not trip on a cat or disturb the neighbours for nothing.
The goal is not to collect videos of attempts, but for there to be no attempt. A perimeter that reacts in time changes the equation for whoever is watching from outside.
Architecture
Why no single layer works alone
Every technology has a blind spot. Radar covers volume but not fine detail; LiDAR measures a plane precisely but not the whole garden; thermal sees heat at night but does not decide what it is; AI decides but needs something to look. Selling any one of them as a complete solution is selling a hole.
The architecture that works in your area combines all four in layers: radar sweeps the open volume, LiDAR seals the critical planes, thermal covers the unlit lines and AI provides the judgement that says whether a detection is a person. The alert only fires when two layers agree, which is why it does not trip on a branch but does on a real intruder.
How many layers and where depends on your plot, not on a boxed package. That is why everything goes through the visit first.
Access points
The weak point isn't the fence — it's the gate
In most homes the intruder does not climb the highest wall; they come in where there is already a gap made for coming in: the vehicle gate, the service door, the gardener's or the pool access. These are the points legitimate people use every day, and precisely for that reason the worst-watched.
In your area we treat those accesses as zones in their own right: detection that tells the expected passage from the one that is not, image verification of who arrives, and a record of every opening. The point is not to turn your home into a bunker, but to know who comes in when you are away, and to make sure an open service door is not an invisible hole.
Where there is domestic staff, maintenance or frequent deliveries, this is the layer that turns a normal flow of people into something traceable, with no friction for you.
Shall we talk about your home?
No-obligation technical visit anywhere in the province of Malaga. We work in English and Spanish.