Solutions by typeIntrusion detection
What if a camera stops seeing and nobody notices?
A camera that has stopped seeing
It’s 7:40 p.m. A cell landing seen from a high camera. In the right foreground, someone with their arms up throws a garment over the lens, right next to the rules board. The frame turns into a smear of gray fabric. The recording doesn’t stop: it keeps recording, and it records a rag perfectly. In the control room, the sixteen-camera wall still has sixteen tiles.
1 cameras across 1 sectors
This is how you ask for it
«Tell me the moment a camera stops seeing: if it is covered, if it is turned, or if the frame suddenly goes black.»
A covered camera doesn’t raise an alarm: it stops raising them. And on a screen, that looks a lot like everything being fine. It’s discovered weeks later, when someone goes looking for the video of an incident and finds an hour of gray. By then it isn’t only that video missing: it’s the confidence in all the others, because nobody knows how many times the same thing happened on a camera where nothing was ever looked for.
You decide once
You decide this once. You say which cameras are critical, how many minutes without a usable image are tolerated before it alerts, whether the scheduled cleaning window is exempt, and who receives it: usually maintenance, not security, because the person who fixes it is the one with the ladder. And you also decide who else has to know: the person who wrote the rules that hung off that camera, because while it’s blind those rules aren’t being applied.
This is the only word in this vocabulary that isn’t about what happens in front of a camera, but about the camera itself. And it’s the easiest one to forget, because it produces nothing: no alert, no incident, not a line in any report. It produces silence. And silence, seen from a control room with sixteen tiles lit up, looks an awful lot like calm.
On a cell landing somebody covers it deliberately, and usually right before something happens. But most cameras that stop seeing are touched by nobody. A spider’s web in front of the infrared, giving a white flare at night. A branch that grows all summer until it fills half the image. A sticker. A truck parked opposite for three weeks. Salt on the dome of a port camera. A pane that fogs at six in the morning and clears at nine, exactly across the window that matters. And the worst of all, because you can’t see it: the installer who knocked the housing while climbing the ladder and left the camera looking six feet to the left of where it should. They all cost the same: the hours in which that camera was worth nothing.
And there’s a second effect, worse than losing that video, that’s almost never mentioned. Every rule that hung off that camera has stopped being applied, and it stopped without saying so. The lone worker in that room is no longer being counted. The counting line on that door no longer adds up, and the monthly figure comes out lower with nobody knowing why. The stretch of perimeter that lens watched no longer exists. A blind camera doesn’t warn you that it has stopped warning you: that’s the quietest failure a video installation can have, and the only one that gets worse the more you rely on it.
That’s why this rule is usually not asked for by security: it’s asked for by maintenance. IRIS looks at what the camera is sending and says whether there’s a scene there or a smear. When the scene goes, someone with a ladder fixes it on the same shift and the rules come back to life. It isn’t the flashiest thing IRIS does. In any given year, it’s the thing that most often prevents an uncomfortable conversation.
What changes
Before, it was discovered when someone needed the video, and by then it was too late by definition: you can’t recover an hour that wasn’t recorded properly. Now it’s discovered within minutes and fixed the same day, by the person who always fixes it. And you get the record nobody had: which camera, how long it was blind, and which rules depended on it during that time. With that, you can have a serious conversation about the installation.
Where we have seen it
CAM-113The camera has been covered
«Tell me the moment a camera stops seeing: if it is covered, if it is turned, or if the frame goes black all at once.»
What it doesn’t do
IRIS does not know why a camera stopped seeing. It does not tell a garment from a spider’s web, a parked truck or fog. It says this: this camera is not delivering a usable image, and it started at 7:41 p.m.
And if the camera loses the signal entirely, that is no longer a question of image. Your recorder or your network sees it before anyone else. IRIS looks at what arrives; when nothing arrives, it is the system that raises it, not the sight.
Questions
- Can it tell whether it was covered on purpose?
- No, and it doesn’t try to. IRIS doesn’t know whether there’s a garment, a spider or a truck in front of it: it knows there is no scene anymore. And for what this rule is for, that doesn’t matter, because the cost is the same in all three cases: the hours in which that camera was worth nothing. Whoever goes up the ladder sees in two seconds whether it’s a web or something else, and from there the kind of problem gets decided. That part is done by a person.
- What about a camera that slowly drifts out of focus?
- Yes, and that’s the case that gets used most, even if it isn’t the one that sounds best. A dome slowly fogging over a month is seen by nobody, because each day looks like the last and the eye adapts. IRIS compares against what that camera used to deliver, not against a general idea of sharpness, so it notices the decline even when it’s slow. In a large installation, this rule usually finds more broken cameras in the first week than the whole previous year.
- Does the recorder not do this already?
- In part, and the missing part is exactly the one that hurts. The recorder knows whether the camera answers and whether video arrives. It doesn’t know whether that video is any use. A camera turned to face a wall answers perfectly, sends its stream on time and raises no incident at all: it’s healthy and it’s seeing nothing. The same goes for a fogged dome or a salt-covered lens. What IRIS adds is looking at the image, not at the device.
- What happens to the rules that depended on that camera?
- They stop while the camera can’t see, and that’s why the alert can also go to whoever wrote them. It’s the most important thing about this word and the least often said: if that camera was counting the person working alone, during that hour nobody was being counted. If it was adding entries, the day’s figure comes out lower. When the image comes back, the rules resume on their own, and it’s written down how long they were stopped, so nobody later reads a number that isn’t what it looks like.
Demo · IRIS at work
A new case each time. Always the same screen.
It plays by itself and shows one case from start to finish. It’s a summary: the real screen has far more functions than fit in this box.
- Cameras on the map
- You draw an area
- What passed shows up
Not seeing yours here?
These words are the ones we have built, not the ones IRIS understands. The list falls short on purpose: you ask for it in a sentence, and sentences don’t run out. Tell us what is in front of your cameras and we’ll tell you whether it understands it, or not yet.
And the best part
You can add more cameras;you can’t add more hours to anyone’s day.
Cameras can be multiplied; a person’s hours can’t. IRIS looks at everything coming in and leaves them the one thing no machine does better: deciding what to do.
Tell us your problem and we’ll say whether IRIS understands it, or not yet.
A person answers, the same working day.