Something that frustrates me about video surveillance
The number of useless video surveillance installs out there really frustrates me. Take this example of two men stealing a Maserati. They are caught five times on the CCTV system but every image is useless.
This is extremely common place in many CCTV designs and is really easily avoided in most cases with a little education.
Why are the images useless?
The first security camera is covering what looks like an entry door to the carpark. To recognise a person with this camera resolution (4CIF) – before recording compression – we would need to shrink the area of coverage to the red square.
Now I realise that this would most likely not have done any good as they would be out of the scene at the point of recognition but that is the reality of a 4CIF camera. The camera also has a very poor dynamic range – the ability to manage both light and dark areas of the scene or contrast.
Was it intended to watch the door AND the footpath? Well they needed two cameras or a megapixel then – end of argument.
In image 2 we have much the same issue. The scene is too wide to recognise a person and a big problem with the dynamic range. Additional the camera appears to need a good clean.
A choice needs to be made. Do you want the CCTV to recognise people walking in from the foot path – the red square – or get some detail of the car – the blue square. Currently this image is only good to prove if the person you ran over was in the wrong or not.
Image 3 is fine for the scene width to recognise a known person but again the dynamic range of the camera is letting us down. The easy fix outside of a betted camera would be to swing the camera to the right so we are looking in to the darker areas only. Better still – more the camera to look directly at the door face on and fill the hight of the image with the door. A perfect identification image.
The last two images are of the car leaving. The car, while still inside the carpark is blurred because the camera has it’s shutter speed too low. This is common for low light images but it should be at around 50-125 for a moving vehicle, speed dependant. The car entering the road would give a great action shot if the car had crashed but not much good for anything else.
To summerise
Having put down each camera here, I need to add that carparks are not easy to cover effectively.
The key to good coverage is to have an overview of the roads as per the blurred image above – but with out the blurring 
– combined with registration cameras at the entry and exit points. This is typically going to be a maximum of 4 more cameras.
Having now covered all internal vehicle movement and registration on entry exit we are able to determine in most locations, which car damaged another vehicle by following it on the road way overviews and then identifying it by it’s numberplate as it leaves.
For the people walking in the car park, the registration cameras MAY get a good face shot but you will need other cameras on the main stairwell exits, lifts and street access points.
Yes it will cost more
But why bother spending money on a CCTV system that only does a quarter of the job. If you paid $20,000 for the above system – assuming there are other cameras as well, an extra $5,000 would have completed the job.
