frank3

vs Looks the same to me.

I heard the Avigilon encoder only works with Monster Cable high performance Ethernet cables. http://www.monstercable.com/productdisplay.asp?pin=4697

3% CPU for 23 cams?? How much bandwidth is that server dealing with? Are you sure the cameras are connected?

Check out Avigilon Standard. http://www.avigilon.com/products/controlcenter/overview/ It's very cost-effective even with non-Avigilon cameras. It lacks some features but it's just as powerful as their Enterprise edition. The best part is their software is a one-time cost, there is no license renewal and I've never had to pay for support, software updates and new features.

It's my understanding that JPEG2000 allows much higher quality than anything else because it supports the full quality of the camera sensor. H.264 is limited to 8-bit (256 color shades), whereas Avigilon is using something like 14-bit (16384 color shades) according to some YouTube video. It's a pretty amazing difference. The fact that it allows monitoring multi-megapixel cameras using the bandwidth of a tiny analog camera is also very nice in my opinion. The JPEG2000 file size is big initially but you never really have to pay the price for it, because of smart streaming and data aging.

The megapixel cameras I posted samples of aren't supported by GeoVision so you couldn't work your way up without eventually replacing it. The Avigilon Pro cameras only work with the Avigilon NVR. Fortunately, that NVR supports analog PTZ cameras very well through inexpensive video encoders converting 4 channels into IP cameras. If you want to build your system progressively all the way to the very best quality, then you should go with a solid PC, a few terabyte hard drives and 4 network encoders (16 channels), then start adding megapixel IP cameras (Avigilon or any other supported megapixel brand) and eventually move up to the 16 megapixel Avigilon Pro cameras where the extra quality and resolution are needed.

Here's a few Avigilon megapixel nightshots, courtesy of megapixelman's photobucket. Pro daytime for comparison : Pro color (mono 16MP is even more sensitive at night) : Non-pro : How does that compare to CNB?

I agree with others that you should investigate Avigilon. They have a unique and very scalable, flexible offering. For any system requiring outdoor surveillance their high-end 16 megapixel cameras are just fantastic, yet their economical 1080p H.264 cameras integrate seamlessly in the whole design. They're also Canadian so I suppose their cameras are able to withstand extreme winter. See their online case studies, they designed a lot of college campus systems : http://avigilon.com/products/documents/casestudies/ I'll try to answer some of your questions... As far as I know all their cameras have built-in motion detection to trigger recording. They don't have internal recording but their HDSM technology works great across wireless links or simply in a distributed architecture across many buildings so you can spread the NVR servers around. Their cameras are day/night and become monochrome (infrared-sensitive) at night. They offer rugged outdoor domes with built-in heaters for the 1-5 MP cameras and outdoor enclosures for the 11-16MP cameras. Third-party integration is quite decent too so you can add many other brands later on, including PTZ cameras where useful. It also integrates with a bunch of access control systems. Searchability is amazing with their pixel search feature, letting you go through days of footage in seconds. I don't know how it works, but it's just instant! Maybe thewireguys can comment on that?

This looks like a great little camera and it will most likely be below $2000 since they use the same body/lens type with smaller CMOS pixels. I'll definitely buy one for daylight events. I wonder how well it will resolve such a resolution with a small lens, though. The Avigilon camera you mention is expensive because it uses a huge 35mm sensor with the type of lens you would install on a high-end DSLR camera. That camera can capture amazing detail in a stadium concert with the lights off. In many scenarios, it's better than hundreds of cameras that can't see.

Recording is definitely not a bottleneck. A terabyte hard drive is dirt cheap nowadays and those cameras can be set to only record motion, so it's as effective as H.264 in terms of storage. On top of that, the flexibility of JPEG 2000 allows a same stream to age over time. Whereas with H.264 we might be stuck with 10 days for 1080p 30fps (unless we reduce the frame rate upfront), the Avigilon NVR can keep the stream over a year if needed, since it can peel frame rate, quality and resolution layers like you would peel an onion. We configured our NVRs to keep at least a month of full resolution and a year of progressively aged footage. The same principle applies to recorded footage, not just live. It's possible to remotely monitor live cameras over the Internet, yet analyze past footage at the same time, on dozens of cameras. One good example is their thumbnail search feature, which lets you instantly view hundreds of thumbnails for the last hour, the last day or even the whole month. From my understanding, this would be technically impossible with H.264, since you would need to download lots of 1080p pictures, a bandwidth-intensive operation. My point is, for most useful uses of a video surveillance system, megapixel is not needed until you absolutely need to zoom in into details, but other codecs force you to absorb and handle all the megapixels, whether you need them and can handle them or not. Since in most scenarios we don't, I have found the flexibility of JPEG 2000's on-demand behaviour to be much more suitable and powerful. To answer your last question, the NVR decides which parts of the images to transmit. It doesn't need to down-scale or re-encode, it just sends less data than what it receives from cameras. I don't know the details, but I found a YouTube video that gives a better explanation. See around 3:45, http://www.youtube.com/watch?v=1Fmle3bkSYQ All I know is that it does a darn good job, even with our third-party analog cameras.

I understand from your post that you work for Indigo Vision, so you're unlikely to recommend your competitors. However, for the benefit of others in this forum, you should keep your claims closer to the truth. First of all, JPEG2000 is an ISO standard which is widely used for digital cinema (movie theaters) and military imaging/surveillance. The only reason it's not being used by the TV industry is that there is always a single fullscreen stream and the resolution is fixed in advance. CCTV needs not play catch up with the TV industry, it can do better. We may be an atypical Avigilon customer, but we have lots of SD analog cameras, rather than megapixel. Even then, JPEG2000 has significant benefits for us. It enabled possibilities and solved problems that other surveillance companies won't even attempt to tackle. Let's say you are monitoring 36 analog cameras on a dual monitor setup. That's 0.3 megapixel per camera for a total of 15 megapixels. Each monitor has 1 megapixel of visible surface, for a total of 2. With JPEG2000, the software isn't streaming and processing 15 megapixels, it's only streaming and processing 2 megapixels, which is what you need for full quality on your monitors. Whatever you throw at it on any computer, it will always scale beautifully. With the H.264 systems we've seen, you couldn't even come close to that efficiency, let alone stream those cameras across the Internet simultaneously. Say we had 36 1080p cameras instead. That would be 72 megapixels of data for our 2 megapixels of monitor surface. JPEG2000 still streams 2 megapixels regardless, or progressively lowers frame rate, quality or resolution if we wish to monitor across the Internet. H.264 would stream 72 megapixels at 30 frames per second (2 gigapixels per second), or downgrade to one frame every few seconds since the computer and network couldn't keep up.

By the way, I didn't want to offend by not sharing your excitement, so sorry if I did. After having seen the video, I just wanted you to realize that you wouldn't get such a good quality/value ratio compared to competitive offerings, whether it's equivalent at a lower price or higher end at a similar price. While I admit the 10x optical zoom is interesting, it would only make sense if paired with pan/zoom and a human operator. Without that, it's better to buy a quality lens with a fixed zoom calibrated for the optimal field of view for your purposes. Also, it would be nice if they didn't try to trick people with 1080i versus 1080p, because interlaced footage is definitely horrible when paused. That should really be avoided.

From what I've seen and tried, mainstream license plate analytics require characters that are at least 50 pixels high to achieve decent accuracy, so as long as a character is 1/10th the height of a VGA camera, it should work. With 1.3 megapixel you should be able to have the complete car in your field of view, with a character being 1/20th the height of the picture. Thus I also doubt such a system would be sufficient to detect 3 lanes in an automated fashion, since a full license plate at that resolution would cover approximately 1/5th (20%) of the image width. The best license plate analytics module I've tried could detect characters as small as 10 pixels wide by 15 pixels high, but that was with a camera factory calibrated for that purpose, not a generic IP cam. Even with that kind of accuracy, a single camera was limited to two lanes or else sensitivity would drop. I'd be curious to hear about your experiences and opinions too. I only tried 3 license plate systems so far.

Thanks for the videos. I was curious after reading your claim, but that was enlightening. I agree the first video looks decent. However, it's all about tricking your eye with the motion, like when you watch HDTV. You're not supposed to extract sharp shots out of it. As soon as you hit pause to try and recognize something or somebody, you'll notice a huge problem. Five cars drive by right in front of the camera, and even with proper auto-focus, their plates are unreadable. There's only one out of five that I can guess at, but I'm even not sure about the first letters (small red car). If you look at the flow, it's a good video, but it's not optimal for a security application. You can find a lady with a red bag, but you won't be able to give the black car's plate number to the authorities if you had to. HDTV needs not be the baseline for CCTV, fellas. I'm personally tired of reports saying CCTV is useless when incidents occur because details are needed yet are missing. CCTV can be so much better than that when you avoid expensive marketing fallacies. The second airport video makes the camera look absolutely useless. One megapixel in such a large area is unsuitable except to know whether it's day or night. Even then, I would like to see that footage without the bright sunlight because the camera is having a really hard time with contrast. When the camera is zoomed out, you can't read or recognize anything. Everyone appears to be wearing black clothes due to the poor colour resolution. See how the bright colors spread over a few pixels, preventing light signs to be read. When it zooms in, you lose the whole airport perspective, and while you can differentiate hair colour, all faces look the same. When it zooms to a usable level, the picture is shaky and blurry, and you view such a small area that most people's faces are outside the field of view. After seeing those videos, I can say your comparison with Avigilon was probably misguided. A 16 megapixel camera lets you see so much more without ever compromising field of view. The lack of motion blur is also immensely useful in security applications, because every image exceeds that of a high-end professional digital camera, so all elements are sharp and recognizable. It is true however that the 16 megapixel is a kind of no-compromise product that's in a different pricing league, so it would be fair to compare the Axis to their competing product line instead. The same Axis price would easily get you a 5 megapixel day&night Avigilon camera, which would be far more impressive in an airport with a zoom lens. For the first scene with pedestrians, one megapixel is fine but you should get a camera with a quality sensor so that individual pictures aren't as blurry. Cameras with higher sensitivity and lower exposure times would generate pictures with readable car plates and recognizable faces. I don't want to promote any online reseller, so you will have to look it up, but I've seen one carrying the Avigilon 1MP for less than a third of the price of that Axis camera, or cheaper than any other Axis camera for that matter. Even their 2MP is very inexpensive, and that would give you real 1080p rather than 1080i.

IP cameras are similar to digital cameras that send images over a network cable instead of to a memory card. Images are sent in multiple IP packets tagged with a destination address. You can think of switches/routers as post offices that receive all the packets and forward them through the right cable, because it learns which addresses are accessible trough which cable. If you have an 8-port switch, you can connect 7 cameras with 7 cables, and use the last cable to connect a computer with NVR software. You can also connect multiple switches together, and as long as IP camera packets can follow a path from switch to switch all the way to an NVR, then the NVR will be able to know which camera address it comes from and will be able to record the images. Depending on the bandwidth of your cameras (the sum of all cameras), you may want to use a 100Mbit or 1000Mbit (Gigabit) link to your NVR. Given current prices, it is probably safer to go with Gigabit switches to make sure you have plenty of bandwidth, unless you have very few cameras. An NVR can also be an appliance, but it is usually a normal computer/server with software that handles storage and can allow access by other remote computers. Typically, an NVR is connected to both a camera network and an external network through which client computers can access the recorded data. Hopefully that's helpful!

How can a 1080i look any good with all the interlacing tearing involved? For surveillance applications it matters to get sequences of good still shots, not videos with motion artifacts. You would be better off using that camera in 720p progressive mode, but then that's only 1 megapixel. Why do you say it's a game changer? I'd like to see your demo video though. Is it available online?

That's a sweet little design. I usually did fixed IR but automatic sounds nice. Did you get it to work? Where did you buy that filter mount? I would like to try on some cameras.

Do your cameras by themselves look good if hooked up directly to a TV? If not then there's not much you can do as that's your source image. You may be able to get better night time performance by removing the infrared filter under the lens, though. Otherwise if they do look good, you may want upgrade your image capture device. As I was suggesting another folk on this forum, the Avigilon networked video encoder is thus far the best I've found. It's a little box that turns 4 analog cameras into IP cameras which you can access on a PC. This allows you to transmit images over longer distances in digital format to avoid signal loss, too. From my experience it makes all analog cameras look very sharp on monitors. You will need a PC and a single-camera software license though, but with the encoder it's still cheaper than a standalone DVR. Hope that helps getting the most out of it!

Did you try resetting the camera or warming up the sensor? I've seen this happen on cheap cameras with wide temperature gradients, once it gets cold.

If you want to get the most of your analog cameras, you should look at Avigilon's networked video encoder. It's much cheaper than the typical DVR card (a few hundred dollars), and it let me record 4 D1 channels (up to 768x576) at 30 fps with no drop in picture quality. In fact it looks much sharper on the computer than it did directly on my TV. I think they do adaptive contrast enhancement. Basically, it turns 4 analog cameras into individually accessible IP cameras with smart motion-triggered recording. Since those 4 channels count as 1 IP camera, it only takes a single camera license on the Avigilon software, so it's definitely the most cost-effective solution I've found so far.

After reading about Avigilon on this forum, we recently deployed our first Avigilon system and are totally blown away. It's a totally different league from the IP cameras we had before, and we unfortunately tried a lot of different manufacturers. We mostly went with their affordable line of cameras, but if you say money is not an issue, you should definitively give their Professional line a try. They're a bit more expensive but totally worth every single penny. I think they're the only guys out there using high-end scientific CCD sensors with image compression that meets military standards. If you want to identify anything at night, you'll want some of those.

Have you managed to find any? I'd love to see a photo gallery of sample images. I found an example which is quite impressive. I can't post the address, so google "avigilon photoshop" and you'll find Forensic Photoshop.