WestPoint · Camera systems
There is a camera pointing at a hedge and has been for two years.
And nobody knows, because recording is not the same as seeing. Complete CCTV: project, equipment, installation, commissioning and maintenance — with the field of view decided before anyone climbs the pole and checked at every inspection.
Nobody looks at what they record.
It is the awkward line about any camera installation, and it is worth starting with it.
In most of the places we walk into there are already cameras, and they work: the recorder has disc space, the images arrive, the reception monitor cycles through mosaics. What there is not is anyone able to say what can be seen from each one, because since installation day nobody has checked again.
So three things live there unnoticed. A camera that has spent two years looking at a hedge that grew. Another that faces straight into the sun at six in the evening and for forty minutes is a white smear. And a third that is perfectly framed but records with so little detail that the face you need comes out at twelve pixels.
All three turn up on the same day: when an image has to be pulled out. And on that day nothing can be fixed, because what was not recorded does not exist. Nobody complains about cameras until they are needed, and that is exactly why they have to be installed with that day in mind.
What are you going to ask of that image.
The whole installation comes out of this question, and it has to be answered camera by camera before buying any of them.
A camera is not there «to keep an eye on things». It is there for one of these: to see that somebody is there, to know what they are doing, to recognise a person you already know, to identify a stranger or to read a number plate. Each needs more detail than the last on the same scene, and the detail gets shared out: a camera covering twenty metres of width cannot have the same detail on a face as one covering three.
Everything else follows from that. How much scene each camera takes in, what lens it carries, how far it is from the place where the thing that matters happens, at what height and at what tilt. Ten degrees too many and it stops reading a number plate and starts seeing the roof of the car, and no software fixes that afterwards.
And there is one cut that always gets made and almost never said out loud: spreading the money that would have bought twelve well-placed cameras across twenty. Twenty cameras that do not reach the detail you needed do not cover more than the twelve. They cover the same, worse, and with twenty places to fail and twenty fields of view to check.
Night, sun and rain.
Cameras get chosen by looking at a catalogue image taken at midday. Afterwards they work at night, in the rain, in fog, into the sun and with a car's headlights coming straight at them. Almost every problem with a CCTV installation is there, not in the camera.
Against light coming straight at it there is equipment that compensates fairly well, and there are places where the only thing that works is moving the camera, because nothing compensates for a direct setting sun. At night, the question is not whether the camera «sees in the dark»: it is whether there is enough light, where it comes from, whether it is ours or the street lamp's, and what happens the day the neighbour turns theirs off.
All of that is checked beforehand, on the visit, at the bad hour. It is the difference between handing over an installation and handing over an installation that works on the days you need it.
What gets decided about the recording.
The part that does not get discussed in the meeting and the one that causes the most trouble later.
- How long it is kept. It is not a technical number: it depends on how quickly you find out about things. If a stock discrepancy is spotted at inventory time, fifteen days of recording is no use at all.
- What happens when a drive breaks. They break. The question is whether the system raises an alert and carries on recording, or whether somebody finds out three weeks later when they go looking for an image that is no longer there.
- Who can see what, and who can delete. The cameras also record the workforce, and that is the most sensitive thing there will be in the installation. There have to be named users and a trace of who has looked and who has exported.
- How an image gets out of there. The day a video has to be handed to the police or to an insurer, it has to be able to come out in a format that plays elsewhere and with evidence that it has not been altered. If exporting means ringing the installer, the system is not yours.
- Where the recorder is. Recording equipment inside the same premises being watched is the first thing to be taken. It is a decision about place, not about brand.
And one thing only we can do here.
Mounting the camera knowing what will be asked of that image in two years' time.
Infinity Neural is in the same house, and it is the part of the group that makes a camera understand what it sees: counting people, alerting that somebody has entered an area without a hard hat, reading number plates, measuring how long a loading bay has been occupied. We install the cameras and they work with those images, so we know at first hand what kind of image is useful to them and what kind is not.
That changes something small in the project and large afterwards: when a camera is fitted knowing that one day it will have to count or read something, it goes where it needs to go from day one. The angle, the height and the light are the same ones that would be needed in two years' time.
It is not compulsory and most of our installations do not have it. But it leaves the door open at no cost, which is the opposite of finding out later that analysing means going back up the poles and paying for the installation twice.
Infinity Neural
What makes a camera understand what it sees, tuned to your particular scene: your angle, your light, the way you work. Here is what an image can be asked for and what it cannot.
IRIS Neural
The other way: a system that is installed and covers the whole site, with the video, the rules and the alerts in one place. It works with the internet cable unplugged.
What gets handed over and what gets tested.
A camera installation is handed over with a sheet per camera. Anything else is word of mouth.
- A sheet per camera: where it is, what can be seen from there, what it is for, what it does not cover — and its daytime image and its night image, taken on the day of handover. That sheet is what lets you check in a year's time whether the camera still sees the same thing.
- The recording calculation: how many days are kept with the configuration handed over, and what happens to that number if five cameras are added tomorrow.
- The details of every piece of equipment: brand, model, serial number, address and password, in the client's name.
- Training for whoever is going to use it: searching, exporting and, above all, what to do when something does not turn up.
- The inspection plan, with what gets checked at each one. And what matters most: that at every inspection the field of view of each camera is looked at again, because hedges grow, lorries start parking in new places and industrial units get reorganised without telling anybody.
What we do not do.
We do not give a price per camera without having seen the site. Anyone who does is putting the same camera at all twenty points, and in a real installation no two points are alike.
We do not promise that a face or a number plate can be pulled out of any image. It depends on the detail that reaches that particular point, and it is said to your face, with the image up, before signing and not on the day of the incident.
We do not tie ourselves to one manufacturer. We work with many and the choice is made on what is needed at each point. When a supplier always fits the same brand, the project looks suspiciously like their catalogue.
And we do not invoice maintenance that does not get done. A camera inspection that does not include checking again what is visible from each one is a visit, not an inspection, and it is the most common way of charging for nothing in this trade.
The other seven areas.
Cameras almost never work alone: they verify the alarm, they document an entry, and on the day of a fire they are the eyes of whoever is deciding.
The brands we build with.
We belong to none of them. An integrator who only knows how to fit one brand is a distributor in overalls, and what decides which camera goes at a point is THE POINT: the light there at six in the evening, the distance, the backlight, whether a number plate has to be read or seeing that somebody went past is enough. Each one with where it behaves well and where we would not put it.
What people ask about cameras.
The ones that come up most. There are 40 on this subject, and the other 34 are in the question index.
All the questions, by subject▸Dome or bullet? Which one suits me?
The place decides, not taste: the dome goes where the camera is within arm's reach or where you do not want it obvious which way it is looking, and the bullet goes where you have to reach a long way, with a long lens and a shroud to protect it from the sun and from driving rain.
The real difference is the bubble. A dome puts a polycarbonate dome in front of the lens, and that dome gets dirty, reflects and, with the infrared on, throws light back into the lens itself if the internal sealing is not good. In exchange, it is harder to knock out of position and birds do not perch on it. A bullet has no dome, so the image comes out cleaner at night, but it is a tube aimed from outside, and at head height anybody can move it.
There is a third case that gets forgotten: in high, clean interiors, the dome hangs from the ceiling and takes up no wall; on a façade, a bullet with a shroud stops water running down the glass and leaving streaks that show up on every image taken in the rain.
▸What is a multisensor camera for and when does it pay?
It is for covering a wide area with uniform detail from a single mounting point: it is several independent lenses in the same housing, each with its own field of view, presented to the system as several cameras even though there is one cable and one power feed.
It pays when the place to hang it is expensive and the scene is wide. The corner of an industrial unit, a junction of corridors, a car park canopy: where you would have to run three cables to three different points, here you run one. And it pays when the width is at different distances, because each sensor can carry a different lens and one can look at ten metres and another at thirty with the same detail on the ground.
It does not pay when what you have is a narrow scene and a single direction of interest: there you are paying for sensors looking at a wall. And you have to check the software licence before buying, because many systems count each sensor as a camera and what you save on cable you pay in licences.
Coverage: how many cameras and from where Engineering: the project before buying
▸Do I need a PTZ or will fixed cameras do?
A PTZ is only genuinely useful if there is somebody moving it, or something that moves it. A motorised camera nobody touches is looking at a fixed place with the added problem that nobody knows which one, because the last operator left it wherever they left it.
The two cases where it makes sense are clear. One: there is a control room with people on shifts following someone across the site. Two: it is tied to something else —a fixed camera that detects, an access reader, an alarm— and swings by itself to a preset position when the event happens, which is how most of the ones that work well operate.
And there is a detail you do not see when buying: when the PTZ is looking north, the south is not being recorded. That is why on a site where having the full history matters, the PTZ goes in as well as the fixed cameras and not instead of them. It is also the camera with moving parts, and the motor and the slip ring are components that wear out over the years.
Types of camera and what they are for The operator position, running in IRIS
▸When do you need a thermal camera?
When what you have to do is detect presence in total darkness or on a long perimeter, and not identify anyone. A thermal camera does not see light: it sees differences in temperature, so it works just the same at four in the morning with no street lamp, in light fog and with the intruder dressed in camouflage, and it is not blinded by a car's headlights.
Its limit is that it does not give a face, a number plate or the colour of clothing. A white outline on a dark background does not identify anybody to a judge. That is why the usual pairing is a thermal camera to raise the alert at a long distance and a visible-spectrum camera with light that comes on, or with infrared, to get the image that is useful afterwards.
It is an expensive camera for the number of pixels it gives, so it goes where there is no alternative: long unlit fence lines, ponds and tanks, open yards of material. In a yard with street lighting, a well-chosen ordinary camera usually solves the same thing for less.
Types of camera and what they are for FLIR: thermal, what it is for and what it is not Detection over the scene, running in IRIS
▸Is a fisheye camera any good for covering a whole room?
It is good for seeing everything that happens in the room and knowing where each person moved, and no good for recognising anybody away from the centre. The sensor spreads its pixels over three hundred and sixty degrees, so the detail at any one point is a fraction of what a camera pointed at that point would give.
The practical rule is ceiling height and useful radius. Hung in the centre of a room with a normal ceiling, what is directly underneath is seen well; beyond a certain radius the person is seen from above, foreshortened and with few pixels on the face. That is why it fits square, medium-sized rooms very well —a shop, an open-plan office, a lobby— and long narrow industrial units badly.
Almost all of them correct the distortion in software, either in the camera itself or in the video software. It is worth checking where that correction happens before buying, because if it is only done at the viewing workstation, exported video can come out as the uncorrected circular image, and that is what whoever opens it elsewhere will see.
▸How many cameras does it take to cover a yard?
It comes out of the detail you need and the width each camera takes in, and that calculation is done before buying anything. A camera covers a width of scene that depends on its lens and on the distance; the sensor's pixels are spread across that width; and from that come the pixels per metre reaching each point. If faces are needed at the staff entrance and seeing that somebody is there is enough on the lorry apron, it is not the same number of cameras even though it is the same yard.
The calculation is done zone by zone, not by total square metres. In an ordinary yard there are usually three different things: the entrances, where face or number plate detail is needed and the cameras go low and narrow; the perimeter, where detecting is enough and the cameras go high and wide; and the points where things actually happen —the loading bay, the waste area, where the forklifts park—, which get a camera of their own.
What does not work is spreading the money that would have bought twelve well-placed cameras across twenty. Twenty cameras that do not reach the detail you needed cover the same, worse, and with twenty fields of view to check.
Coverage: how many cameras and from where Engineering: the project before buying
Get started
What can you see from your cameras?
If you already have an installation, the first useful conversation is looking together at what can really be seen from each point and what cannot. If it has to be built, tell us what has to be protected and what you would like to be able to get out of those images on the day you need it.