Is IP camera performance actually up to the job?

Perhaps I’m greedy, but I just want it all!

Do you ever get that thing where you’re standing around the coffee pot or the water cooler having the old *analogue v IP* discussion (no?…okay, maybe that’s just me…) and somebody says “making that comparison is like comparing apples and oranges”?

I think what they mean is that to make such a comparison is like comparing two things that are *completely* different. Apples and oranges actually have a great deal in common…

This is a bit of a post script to my earlier article Resolution or Revolution on megapixel cameras and how much influence they have on people’s move to IP CCTV. It’s also perhaps a little bit of a rant, but I’ll try to keep the anger to a minimum.

I suppose it’s almost inevitable that a lot of the people who are out there selling IP video equipment are IT and networking sales guys who’ve moved over into the CCTV space opportunistically (I had a new distributor in talking to me today out of that exact same category). That’s no problem.

It’s been said (and I don’t entirely disagree) that it’s easier to take the average IT/networking guy and teach him all he needs to know about security than it is to take the average security guy and bring him up to speed on networking. But you know, when it comes to CCTV, there’s…*stuff*…special *stuff*…

I’ve heard a lot of IP camera salespeople tell me that I “won’t believe how good the pictures are” out of such-and-such a camera, only to *see* the actual images and be totally underwhelmed. Perhaps it comes from having webcams as their only point of reference…

Whether the image from a camera is *good* or not can sometimes be a little subjective on the surface, but there are things you get to know when you’ve looked at thousands of images from thousands of cameras, and you get to know things that maybe aren’t all that obvious. Your expectations are raised, perhaps. Showing me an image of some guy standing in a showroom just isn’t going to get me excited. Zooming in on a vehicle plate in the car park with a 35x zoom lens from 20m away in broad daylight isn’t going to make me go weak at the knees. These seem to be the areas the guys who’ve been selling software all these years just don’t seem to be able to grasp.

It’s taken us years and years to get to the point where we’ve begun to get analogue cameras that really perform the way we’d like them to. Different manufacturers have faded in and out of focus as the time has passed, coming up with new gimmicky names for features that didn’t really do anything particularly useful for us, usually enhancing one particular set of conditions to the detriment of some other set of conditions.

True wide dynamic range cameras with good DSPs and low noise low-light capabilities are actually getting pretty good now. We’ve had excellent results from the Bosch Dinion-XF® range – both the box cameras and the static dome versions. I’m disappointed that the same technology hasn’t made it into their speed domes yet though.

By the way, don’t get caught out with the part numbering on these products. It’s not as straightforward as you’d think. The LTC485-xx is the colour/mono box camera and the LTC495-xx is the true day/night. The xx portion of the part number just selects PAL/NTSC and the power options.

But when you go over to the FlexiDomes you have two different versions of the colour/mono (VDC-485-V03 and VDC-485-V04) with the V04 performing quite considerably better than the V03 in low light. The VDC-495-V04 is the true day/night version. There’s only one version of that. It made me look twice at the data sheet and scratch my head a little.

These cameras are great. I did a lot of trialling with them to try to get optimal settings to catch vehicle number plates in the dark with a reasonable degree of success. One thing I would like to see on them is some sort of *event driven profile setting*. The problem was that to get good night time images I’d wind up the gain, trying to keep the shutter speed as fast as possible. But then in daylight the images would be a little grainy and washed out. It would be good if you could set up a couple of profiles and have the camera switch between day and night profile when it switched the IR filter in and out…just a thought Mr Bosch, if you’re listening…

Back to the plot.

There’s an IP version of these which is *exactly* the same camera but with an RJ45 on the back (the BNC is there too – thank goodness!! Even the part number is almost identical). We’re using these a lot too. The IP version is actually even lower noise in low light, because what little noise you get is below the change threshold for the encoder, so it doesn’t bother encoding those black parts of the image where you’d normally see speckles of noise. Excellent!

As I say, these are great cameras, and they set the bar for where I’d like to see camera performance for external applications – however, they’re still not progressive scan, they’re only 4CIF resolution and I *have* seen a few issues in low light at high frame rates with encoding remnants showing up in the image when you get a bright, fast moving object in the field of view. So, as good a camera as it is, it’s still not *perfect*.

Then we come to the megapixel offerings that the guys from the distributors tell me are “the best pictures you’re going to see anywhere…”.

Hmm…

Okay, so we’re down at maybe 10 frames per second max for 3Mpixel in MPEG4. Low light performance is next to terrible (Mobotix state they can work down to 0.1lux…but only with a 1/60^th second shutter speed, so you can’t do 10 frames per second and anything that moves will blur. You can get 0.005lux with a 1 second exposure. Gulp).

So we’re sacrificing the ability to actually get a usable picture at any time of the day and night for the ability to get fabulous pictures in good light, and nothing for the rest of the time?

Probably the camera manufacturers will tell me I need to look at supplemental lighting – well, duhh!! Yes, of course you can put up more lighting but that’s a whole other subject that I’m aiming to cover another day – but for many people the fact that they need to spend all that money on supplemental lighting will be another good reason why they won’t go for megapixel IP.

A friend of mine is an old-school installer. His father started the family business installing alarms back in the 1950s and this guy has been putting in alarms and CCTV for years. He was asking my advice on a system he wanted to install. He’d told his client that IP was the way to go, because the images were going to be *so good*, but when he showed me the site I realised that there wasn’t a megapixel camera on the market that was going to perform well enough to give him the pictures he was looking for; and the best he was probably going to manage would be to use something like Bosch Dinion-XF® cameras recording at 4CIF…the same sort of recording resolution he’d achieve with a DVR and analogue versions of the cameras for less money.

I’d present this as another case in which some security technology salespeople have over-promised, creating a false perception of the state of the art.

Freddie Mercury said “I want it all!”. He also said “it’s a kind of magic!”.

Tags: Bosch, IP CCTV, megapixel, mobotix

General comment | RSS 2.0 | Respond | Trackback |