Playing around with HD surveillance
I wrote in December of 2007 about IQinVision's online camera/lens calculator and the concept of pixels per foot. At the time, I was fairly enamored of the idea, since it seemed like an easy way to describe the benefits of increased resolution and information gathering and I like doing simple math and feeling like I've accomplished something. In the meantime, however, I've had a hard time wrapping my head around part of the concept: How far away to you measure the foot? And if you need a different lens, an optical zoom, to make a certain pixels per foot happen, why not just go with lower megapixels and a better lens - wouldn't that be cheaper? But the online calculators are helping better understand this idea of pixels per foot, and there's talk that the UK's Home Office has established 40 pixels per foot as the benchmark for facial identification in an image, even though in their latest treatise on setting up video surveillance there's no mention of that standard and they continue to spend a great deal of time on "analogue" and TV lines when they talk about resolution at all. So, as I've come to understand it, it's all about "pixels on target" and wherever that target is in the scene how many pixels cover it. If car is 10 feet long, and it's so far away in the image that even though you've got 16 megapixels only 500 pixels actually cover the car, you've got 500 divided by 10 = 50 pixels per foot and that's good enough resolution to figure out the make and model, maybe. There's talk, however, that 80 pixels per foot allows you to reliably read the license plate, and in that case, you're screwed in this particular scenario. Anyway, I had a long conversation yesterday with Avigilon's Dave Tynan and he helped me get a better understanding of designing HD video surveillance systems, and he pointed me to their own calculator tool (and this one is a particularly smart use of the Internets, since they ask you for some demographic information before you use it, and it's therefore also a lead generator). So, I entered some phony information and it didn't allow me in, then entered my real information and it did. I saw SalesForce.com flash up for a second, so maybe you have to already be verified in some way? It's hard to say. Anyway, I played around with it a bit after getting in and here's what I got: If my camera is 30 feet up in the air on the top of a light pole in a parking lot, and I want to read the license plate of a car 50 feet away, and my camera is pointing down at a 45 degree angle, my field of view width will be 48.3 feet, apparently. In order to have 40 pixels per foot on my thing I'm looking at 50 feet away, I would need a 2 megapixel camera fitted with a 17 mm lens. However, if I go over to the IQinVision calculator, which doesn't have a camera height option, and I plug in 48.3 feet for the width I want to cover at 40 pixels per foot, it tells me I need two 2 megapixel cameras. It says I could only get away with 1 camera if I only wanted to see 40 feet away. But I want 50 feet. How come the Avigilon and IQ calculators don't seem to match up? Let's try another example. According to Avigilon, if my camera is 40 feet high, and the thing is 100 feet away, and I want to cover 100 feet of width, my camera needs to be at a 50-degree angle. And, to get 40 pixels per foot, I need just one 11 megapixel camera with a 39 mm lens (this is impressive to me, if true, that I could get away with just one camera in this instance). For IQinVision, it says I need two 5 megapixels (which makes sense), but it seems to indicate I could only go out to about 64 feet and still get 40 pixels per foot for each of those 5 megapixel deals. That may jibe, actually. Anyway, the question in this scenario is: Is 1 11 megapixel camera cheaper than two 5 megapixel cameras, especially considering the extra cabling and time needed to hang the cameras. I don't actually know because of the prices of these cameras isn't exactly readily available to me, but you can start to do some back of the envelope stuff that I could see would be helpful. I encourage you to play around with the tools, especially if you're skeptical of megapixel and the pixels per foot standard. It's kind of fun.