False alarms: real causes, real fixes

The typical night causes are animals (cat, wild boar), wind-blown vegetation, car headlights and shadows. Each has a different fix; identifying which is yours is the first step.

With image classification and an animal category, and perimeter detection that filters by size (LiDAR) or classifies (radar). The dog's event is recorded but does not alert you.

Radar with the “swaying objects” filter ignores trees and bushes; on cameras, AI analytics discards vegetation movement. It is one of the commonest cases in a Mediterranean garden.

Yes, and they exceed any “pet immunity”. The only serious technical answer is image classification with an animal category, which records the event without waking you.

Because a motion detector reacts to light changes. Radar does not see light (it detects mass), and level-1 AI discards shadows and lighting changes.

On cameras and some detectors, yes: drops and insects in front of the lens. Radar is immune, and analytics rules plus maintenance reduce the rest.

The target is close to zero: a system that cries wolf ends up switched off, and a switched-off system protects nobody. That is why post-installation tuning is part of the job, not an extra.

Often yes: re-siting and adjusting detector heights, applying filters and, if needed, adding an AI analytics appliance on the existing cameras. We start with a survey.

Within limits: it is a declared immunity up to a certain weight (typically up to around 25 kg) and depends critically on mounting height and angle. A bad mounting cancels it.

Almost always the mounting: raising the sensor for looks cancels the geometry that gives it immunity. That is why we document mounting height in the dossier.

Yes: a spider or insect on the lens generates alerts. A classic case, solved with proper analytics and maintenance.

That the alert is only raised when detection coincides with an image confirming it is a person. It is what separates a reliable alarm from one that cries wolf.

It is the classic trap: raising sensitivity multiplies false alarms. The fix is not more sensitivity but better classification — detect and verify separately.

Very few: the agreement of two layers (detection + verification) within a time window eliminates most. The technical logic is on systems.

With a technical survey: we review the event history, the site and the mounting, and tell you in writing what is causing them and what is worth keeping.

The usual causes, by frequency: vegetation moving in the wind, animals (from cats to wild boar), shadows and abrupt light changes, car headlights, heavy rain, and insects sitting on the camera or detector. Each has a different fix — so the first step is identifying which one is yours.

Yes, on two conditions: image classification with an animal category (the event is recorded but raises no alert), and a perimeter layer that filters by size or classifies — LiDAR sets thresholds in centimetres per zone; radar classifies person/vehicle and filters small objects.

Within limits: it is a declared immunity up to a certain weight — typically up to around 25 kg — and it depends critically on mounting height and angle. A wild boar exceeds it by far, and an “aesthetic” mounting at 2 metres cancels the geometry that makes it work at all.

A documented provincial reality: in the veterinary college study of captured urban boar, 40% came from Marbella, 30% from Málaga city, 22.5% from Fuengirola and 7.5% from Benahavís. Neither PIR nor radar solves it alone — the technical answer is image classification with an animal category, which records the event without waking you.

As many as you will tolerate before switching the system off — in other words, the real target is close to zero. A system switched off in frustration protects exactly as much as no system; that is why post-installation tuning is not an extra, it is the job.

Often, yes: correct detector heights and positions, vegetation filters, properly configured analytics rules and, if needed, an AI appliance on top of your existing cameras. We start with a survey and a written report; whatever is good stays.