A camera that only records: what is missing?
An objective is missing. The camera that only records was installed without defining what it needs to show, who should be alerted and what that person does with the alert. To warn you beforehand, it needs the right position and light, person and vehicle detection, an alert with a recipient and a recording that raises an alarm when it stops.
Monday, 8:20 a.m., at the store. The back door was found forced open. The manager opens the recorder, rewinds to the early morning hours and finds the scene: a figure in a cap, backlit by a streetlight, with no face you could recognize. The camera did what it knew how to do. It recorded.
Why is so much camera footage useful only after the problem?
Because the camera was installed to record, not for an objective. Nobody wrote down what it needed to show, whom it should alert and what that person would do with the alert.
Axis, one of the manufacturers ARX works with, sums up the path in four steps: define the use, understand the environment, design with purpose and maintain. The same document points out that even the most expensive camera's image becomes useless if the scene is dark at night, if the camera has been turned away or if the connection has dropped.
None of these problems is solved by swapping the camera for one with more megapixels.
The right camera depends on what you need to see
An international video surveillance standard, IEC 62676-4, separates objectives into levels. The four most commonly used:
- Detect: know that someone is there.
- Observe: count the people and see details, such as clothing.
- Recognize: tell whether it is someone you have seen before.
- Identify: have enough image to say who it is.
Each level demands more detail about the person. That is why a camera covering the entire yard rarely identifies a face at the gate. According to Axis, the difference between an overview camera and an identification camera is not the brand or the model. It is how much of the image falls on the person, the field of view and the angle.
The question that comes before the purchase is simple: at this spot, do I need to know that someone passed, or do I need to know who passed?
Position and light matter more than megapixels
- Height and angle. To identify someone, the angle between the camera and the face must be small. The reference Axis uses is under 20 degrees. A camera mounted too high films the cap and the top of the head.
- Backlight. Entrances, garages and tunnels mix light and dark in the same scene. That is where a person turns into a silhouette. At these spots the camera must handle contrast well (the feature is called WDR) and, preferably, not point toward the bright area.
- Night. A camera's infrared light solves the darkness but changes how things look: a dark shirt can appear light in the image. When detail matters, visible light helps. And it also deters.
Movement is not the same as a person
Standard motion detection triggers on anything that moves: animals, branches, flags, shadows. The phone beeps so often that someone turns the alerts off. From then on, the camera is back to only recording.
Cameras with image analytics classify what moves and alert only when it is a person or a vehicle. It works, under conditions that the manufacturers' documentation makes clear:
- the scene needs enough light;
- the camera cannot look from top to bottom;
- the person must take up a minimum size in the image.
Errors still happen, such as car headlights in heavy rain or an animal the size of a person. That is why tuning is done on site, by day and by night, and not out of the box.
The alert has to reach whoever will act
An alert only prevents something if someone receives it, understands it and knows what to do. Before installing, it is worth answering in writing:
- Who receives the alert at each time of day, including late at night and on weekends?
- What does that person do: look at the image, speak through the loudspeaker, call someone?
- Which areas and times generate an alert, and which only record?
Few alerts, and each one with an owner. That is what separates the camera that alerts from the one that merely records.
Recording must not fail silently
Axis describes a common situation in larger systems: arriving on site and discovering that some cameras had been offline for a long time. Without monitoring, nobody notices until the day the footage is needed.
Three precautions cover most of it:
- The system warns you when something stops. Many recording programs monitor cameras and disks and send an alert when one of them goes down.
- The space is calculated in advance. If the disk is too small for the number of days you want to keep, the oldest recording is erased to make room for the new one.
- Someone checks from time to time. A dirty lens and a worn cable also take away the image. Axis recommends checking at least every six months.
If you would like to look at your cameras with these questions in hand, ARX can do that together with you.
Sources
- Axis Communications. Quality with a purpose: image usability in security. Technical paper on objective, environment, design and maintenance.
- Axis Communications. Pixel density based on IEC 62676-4:2014. Explains the detect, observe, recognize and identify levels.
- Axis Communications. AXIS Object Analytics user manual. Difference between motion detection and person and vehicle classification, with installation conditions.
- Axis Communications. Health monitoring for AXIS Camera Station Pro. Monitoring of cameras, disks and retention time.
Shall we talk?
Tell us what you want to solve. ARX replies clearly, with no obligation.