Choosing the right camera
If you’ve looked at surveillance equipment suppliers you may well be bamboozled by the sheer choice of cameras available – why so many? Which is right for your next application and how to make the right selection.
Gathering useable data in the form of good video recordings depends on a number of factors;
Resolution – a result of the number of pixels the sensor delivers,
Compression – the amount of file size reduction carried out by the codec,
Lighting – no light = no picture, poor light may result in a poor picture,
Field of view – wide angles see more but provide much less detail.
In many cases all of the above influence each other, low light can directly affect the final resolution of the image due to noise and the mechanisms of compression for example.
For the purposes of this article we are going to look at IP cameras, although many of the same principles apply to analogue, resolution would be limited and other factors like cable attenuation come into play which we just don’t want to go into here!
In its simplest form the resolution is the number of pixels on the camera’s sensor device; the more pixels the more resolution we can potentially deliver. We can use the pixel count and field of view to establish probable levels of detail and hence suitability for a particular application.
For IP cameras resolutions will typically range from CIF (352 x 240) to 5mp (2592×1944). For most commercial and high end residential applications 1080p generally offers the ideal solution.
A 1080P camera stream is made up of 1920 x 1080 pixels. The horizontal resolution, (1920), divided by the width of the field of view in metres gives the number of pixels per metre. The higher the pixels per metre value, the greater the image resolution.
As the angle of view is increased (by shorter focal length lenses), the level of detail in any given section of the scene will decrease. As a rough guide 60 pixels per metre is good for general surveillance; seeing people, identifying cars, monitoring traffic. 130 pixels per metre provides more “forensic detail” such as recognising a face or capturing a car number plate. Over 250 pixels per metre will capture high detail such as reading cards at casino tables.
Example – reading a car number plate
Here’s a calculation for a camera in a car park – using a 1080p camera and a scene width of 42m won’t work, the resolution either needs to be increased or the angle of view decreased to 14m;
1920 (pixels) / 42 (metres) = 45.7 pixels per metre alternatively… 1920 (pixels) / 14.2 (metres) = 135 pixels per metre.
As you can see, diagram 1 shows the car number plate clearly because the field of view is narrower, making the calculation result in 135 pixles per metre, compared to the 45.7 pixels per metre seen in diagram 2.
IP cameras all compress video to varying extents in order to ensure efficient use of bandwidth on the network and the recording devices. There are a number of codecs used by the surveillance industry but most use H.264 and/or MJPEG. Cameras allow manual adjustment of the compression levels – low compression maintains more detail but eats bandwidth and hard drives, whilst high compression is more efficient but will degrade detail and image quality. Many cameras enable adpative techniques by either setting the bandwidth (CVAC) or setting the compression (AVC). The file size for each frame changes with complexity, movement and contrast, so either bandwidth changes or compression does. Lastly, frame rates can suffer if there is insufficent bandwidth. Fortunately recording devices do not usually compress the images again so what you see is what you get – recorded, and played back.
Field of View and Lenses
There are three factors that determine the field of view: the size of the image sensor, the lens focal length and the distance to target.
A camera with a 1/2.7” sensor provides a wider field of view than a camera with a 1/2.5” sensor and the longer the lens focal length, the narrower the angle (or field) of view – a 12mm lens will provide a narrower field of view than a 3mm lens.
The two types of lenses used in surveillance are fixed focal length and varifocal. Fixed Focal Length lenses, as you’ve guessed – have a fixed length, e.g. 4mm, 6mm etc. and this cannot be changed. Whereas varifocal lenses can be set anywhere between two set points e.g. 3.3mm – 12mm or 9mm – 22mm, providing options when setting your field of view. There are numerous online calculators to assist with this calculation, but remember widest isn’t always best – you may see “more” but you will see it less clearly.
Most IP cameras are supplied with lenses pre-fitted, so lens selection by F-Stop is not usually an option. Cameras can be bought without lenses – normally referred to a CS mount cameras or “box” cameras these allow any CS mount lens to be used and therefore specific performance dictated. A “fast” lens will have a low F-Stop like F1.2 or F1.0 and will let in up to 4x more light then an F2.0 lens. This is the main reason why your phone camera isn’t very good at night.
Low Light Conditions
There are few applications where light remains constant. An external camera’s performance will be affected by the following factors;
Light level – amount of light reflected to the camera
Sensitivity – the performance of the sensor in the camera and enhancement technology like AGC, Sense Up, Frame integration etc.
Lens Speed – the ability of the lens to pass light measured by the F stop of the lens
In surveillance camera specifications LUX refers to the level of light required for a camera to pick up an image. A camera with a good low-light capability will have a lower LUX rating, for example 0.1 at F1.4.
To put this into context a bright sunny day could be between 10,000 – 100,000 LUX, twilight would be between 1-100 LUX and full moon would create 0.1LUX. So, a camera with a really good LUX rating may still deliver reasonable video under full moon lighting.
Camera performance in low-light conditions can be enhanced by DSP techniques such as Sense Up+ which applies intelligent image signal processing to increase gain without adding noise or motion blur.
Most Infra Red lighting in surveillance is provided by LED’s which emit light at 840nm – the camera sensor is more sensitive to IR than the human eye and can make use of the energy emitted above this wavelength. However infra red is monochromatic light so images will not be in colour. To make use of IR light properly the camera needs to mechanically remove an IR cut filter that sits in front of the sensor (it’s there to enable accurate colour reproduction) these cameras are known as “True Day/Night” cameras.
Certain cameras that have integrated IR illumination LEDs will usually specify their maximum range, but caution must be used when interpreting this as IR reflects differently depending on its target surface and maximum range is not the optimum range. For example, grass has a reflective value of 20%, whereas a car may have a reflective value of 75%.
Another useful tip to remember is: always check the specifications of the IR beam angle – you’ll find that this is often narrower than the field of view the lens may achieve.
So it may still look pretty confusing but the good news is – we manufacturers do all the science so you don’t have to. What you should be doing is matching the performance of the camera to the application to ensure you get the right level of detail in the correct field of view and that it will work at night. Other than that it’s a matter of form factor; internal or external (weatherproof), vandal proof (or resistant) with or without IR etc.