|
||
|
|
|
              |
|
 |
 |
||||||||||||||
|
 |
|
||||||||||||
By William Van Winkle |
||||||||||||||
Stepping up to the high-end of indoor cameras, you might consider IQinvision’s IQ755. This is a 5.0-megapixel day/night unit. At 5.0MP, the camera can capture up to 10 fps. To get 30 fps, you’ll need to drop to the minimum 1.3MP setting. Like a high-end SLR camera, the IQ755 can ship with no lens, power supply, or enclosure with a retail point just over $1,500. The camera features PoE and under 0.05 lux sensitivity thanks to its LIGHTGRABBER II feature set. Like many $1,000 and higher IP cameras, the IQ755 features a CompactFlash slot for on-camera recording (handy if the LAN connection fails), a BNC port for use in legacy situations, and digital pan-tilt-zoom (DPTZ). Normally, we’re not fans of digital zoom features, but once you start talking about multiple megapixels, the ability to zoom in on a face digitally without heavy pixelation and blurring becomes feasible. IQinvision lets users crop out portions of the image, such as the top and bottom bands, to help conserve on bandwidth and storage. Motion detection is done in-camera, and there’s no need for a PC-based NVR application because the IQ755 has one built-in, allowing you to view live video via a Web browser and upload directly to network storage. Note that this upload feature is also found in many less expensive cameras and doesn’t solve the problem of needing to monitor multiple cameras simultaneously. One feature missing from the IQ755 that you’ll find in other high-end indoor/outdoor cams like the Mobotix M12D is a PIR (passive infrared) sensor, which can detect motion in pitch-black conditions. For a brand you may be more familiar with, let’s try D-Link’s top of the line DCS-6620 ($700) and DCS-6620G ($1,000). The only difference between these two cameras is inclusion of 802.11g wireless connectivity. We’re leery of endorsing 2.4 GHz wireless for surveillance because it’s prone to insecure configuration, radio congestion and interference, and, as D-Link technical marketing engineer George Cravens told us, “He who has the loudest frequency wins.” Wireless also greatly increases device complexity. Cravens half-joked with us that D-Link’s first wireless camera Quick Install Guide was 146 pages because of all the different deployment scenarios. Additionally, a full 480p IP stream may only be about 10 Mbps, but when you consider that an 802.11g access point will sustain under 30 Mbps in real life, that means you can only rely on having two wireless cams per AP, which is pretty lame. The forthcoming 802.11n standard (currently Draft N) will remedy many of our technical concerns, but there are no Draft N IP cameras yet.
“Part of the problem is that, currently, Draft N is higher power,” says Cravens, “so PoE doesn’t work so well for Draft N. On top of that, the PTZ cameras use even more power. Even with PoE injectors, you still have a limit on how much power you can provide. A PTZ camera will go from 5V to 12V when active. The IEEE is working on next-gen 30W PoE, but until that happens, don’t expect miracles.” The DCS-6620/G supports up to 704x480 resolution, MJPEG and MPEG-4, still JPEG capture, two-way audio, a 0.05 lux rating, up to 16 cameras on D-Link’s NVR software, and 10X optical pan-tilt-zoom. Neophytes think motorized PTZ is the coolest thing since dry ice, but most experts will tell you that the feature is overhyped. Some IP cameras, such as Panasonic’s KX-HCM280 ($800), feature 21X optical zoom, which is strong enough to let you easily read RAM from 20 feet away. In fact, this sort of fine detail examination is exactly what makes PTZ valuable. The thing is that most usage scenarios don’t call for it. “There’s no hard guideline,” says Axis’ Nilsson, “but in a lot of cases, if you have 40 cameras, four to eight of those will be PTZs—a couple inside and a couple outside. What you have to remember with a PTZ camera is that unless someone is operating it, you’re actually just covering one location unless you have it on a guard tour. But on a guard tour, you can bet that if something is going to happen, it’ll happen when the camera is pointing the wrong direction. In environments where you want to be able to follow someone live, PTZ makes a lot of sense. Otherwise, I’d probably recommend that customers, rather than buy one PTZ, buy three fixed cameras and cover all directions at the same time.” Normally a PTZ camera should be protected and shielded from view by an opaque enclosure (domes are the most common design) so that would-be perpetrators can’t tell where the camera is aimed. The DCS-6620/G is compatible with D-Link’s DCS-70 dome enclosure, which integrates a power supply, heater, and blower. Keep in mind that PTZ cameras entail additional components that may require more regular maintenance and will be more prone to mechanical failure. Going back to the prejudice against PC brands in the security space, we would encourage resellers to actually try out lower cost options before dismissing them. For example, we’ve found the image quality in D-Link’s DCS-5000 and DCS-6600 series to be very close to what you’d find in a more expensive Axis camera with similar specifications. There are limits, of course. Again, below that $200 to $400 line be very careful to test image quality for yourself before recommending it to clients. Even in the home space, we wouldn’t recommend a unit like Linksys’ WVC54GC ($120) on account of its inferior performance under some conditions. “The line is getting a little blurry between good consumer surveillance cameras and the professional models,” says Bob Shouse, senior marketing manager, physical security Specialized Business Unit at Tech Data. “We only carry the consumer line from Panasonic, not their professional models. But for the SMB market, their high-end consumer stuff is excellent. There’s no reason to buy the professional stuff. The consumer line has megapixel resolution and even pan-tilt-zoom.”
Not least of all, look for the little perks that come with the cameras. For instance, D-Link and Panasonic both include free dynamic DNS service for users so they can more easily log into their cameras remotely. (Some vendors charge for DDNS services.) Just about every IP camera—even Linksys’—includes some sort of NVR software. D-Link’s software supports up to 16 cameras. Axis’ bundled app draws the line at 25 cameras, which the company believes is the most streams one server can safely handle, although the software only licenses one camera. Additional Axis licenses are $100 per camera. Lower-end software packages are less likely to integrate with third-party control systems, such as HVAC. Higher-end apps, like Cisco’s Stream Manager 5.0, feature multi-camera synchronized playback, which is useful when you need to review what was happening in several places at the same time. (An interesting harbinger of future IP surveillance: Stream Manager 5 also supports H.264 compression.) More programming-savvy resellers may want to target vendors that offer support for their SDKs in case clients need a higher level of surveillance customization. Video Servers and Switches You might get lucky. You might just walk into an account that has no surveillance equipment and outfit them with IP gear from top to bottom. Even less likely, you might find a client willing to scrap his current analog setup and start from scratch with network cameras. But don’t count on it. A far more plausible scenario is that your customer will already have an investment in analog surveillance that he wants to build on top of with more manageable and modern equipment. This is where we hit video servers. As mentioned above, a video server in the surveillance context is an analog-to-IP converter box. An NTSC or PAL signal comes in from the camera, then runs into an ADC module, much like what’s used on VIVO-enabled consumer graphics cards. The digitized signal then passes through a compression chip, which converts the data into JPEG or MPEG video frames. From there, the stream passes into some sort of CPU-based computer. In the case of a standalone converter card, the card sits in a motherboard slot and offloads data to the CPU. With most video servers, there is an embedded CPU, memory, Ethernet controller, alarm I/O, serial ports, integrated Web server software, and other elements needed to relay a finished video stream on to the NVR. The number and capability of ports on a video server is important, because without the proper I/O you can’t control special camera features, such as PTZ and PIR detection.
Essentially, a video server should in every way convert an analog camera effectively into an IP camera with no loss of functionality. In fact, a good video server can improve on the original camera by introducing everything from image enhancement to motion detection intelligence. Some video servers have special routines that can tell if a camera is being moved or sprayed. Once again, Axis Communications is the clear leader in video server products. The company’s AXIS 24xx product line comes in 1-port or 4-port varieties, and there are a few SKUs that come as bare boards (“blades”) for resellers to integrate as an embedded solution. Axis includes plenty of network security features, event management, a de-interlacing feature for better frame clarity, and support for both IPv4 and IPv6. According to Axis, video servers in SMBs usually get deployed near the cameras. Alternatively, it might make sense in larger convergence scenarios to place video servers in a 19-inch rack arrangement, where they’re likely to be near an NVR server and probably rackmount storage resources. Axis offers two rackmount options. The 291 is a 1U box with 12 BNC ports. There’s also the AXIS Rack Solution, an enclosure able to take up to 12 of Axis’ 240Q, 241Q, or 241S blade cards, meaning up to 48 analog inputs per enclosure.
Still fresh in the physical security convergence space, Cisco is another name to watch. The company’s Video Surveillance IP Gateway Decoder is not only a conventional video server with plenty of I/O, but it has some extra smarts worth noting, such as the ability to output two simultaneous MPEG-4 streams. One stream is full-resolution for recording, the other at low-resolution for live viewing. Cisco’s Decoder also interfaces with existing CCTV monitors as well as many analog CCTV control keyboards and joysticks. “A lot of vendors require you to take the analog cameras, digitize the video, compress it, and stick it in a packet,” says Bob Beliles, senior manager, physical security market management at Cisco Systems. “But then you have to view that live video or record it through a PC. That forces a desktop change in many security operations where the users of these systems have these analog, special purpose control keyboards and joysticks as well as analog monitors. They know these keyboards as well as an accountant knows a numeric keypad in terms of being able to call up video from multiple cameras, specify which display it’s supposed to be on, and then control it if it’s a PTZ camera without ever looking down at the keyboard. To disrupt or change the footprint on that desktop of that operator hits both their potential productivity as well as forcing some additional training. That’s where our decoding capability and preserving these command sets comes in. We don’t see too many other folks with that capability.” Unfortunately, analyst stats don’t break out the opportunity in convergence sales versus regular IP configurations, but we’d wager that for at least the next couple of years, convergence will be the major play in surveillance. Better yet, it’s a move that sets up a lot of potential replacement, upgrade, and service sales for the future. Bandwidth And Storage We can’t advise you on which video management software to adopt. There are too many options, and every IP camera vendor seems to have its own application. Software vendors Milestone Systems (www.milestonesys.com) and On-Net Surveillance Systems, Inc. (www.onssi.com) are two names you’ll likely hear at the start of most NVR software discussions, but there are plenty of others. We’re more concerned with what to do with all of those video streams after they’ve cycled through the monitoring app.
Before we get to discussing storage, we have to address bandwidth. On one hand, vendors like Axis will tell you that bandwidth is no longer a problem. Ethernet throughput has ascended by 10X or 100X while MPEG-4 shaved off 70% of the data load suffered under MJPEG. Unless you’re talking about thousands of cameras in a deployment, 10/100 interfaces at the cameras feeding into a Gigabit Ethernet backbone should handle just about anything. But the network is not where you hit a surveillance bandwidth bottleneck. The real test comes in the NVR server system, both at the disk interface and the central processing resources. There’s still a considerable resource load entailed in handling multiple video streams. A look into the fine print of Axis’ Camera Station management software gives some clues. At CIF resolution (352x240 for NTSC), Axis states that one hard disk can take only 10 cameras at 5 fps. If you want 30 fps in CIF, you’ll probably only be able to stream three cameras to one drive. “A lot of cameras have built-in motion sensing and things like that,” says D-Link’s George Cravens. “You can play with the settings to trade off between frames per second and storage space. If you go too low, the guy’s going to be able to zip right across and you’ll never see him. At best, you’ll have a series of snapshots. But at 30 fps, that’s a lot of data—and data that will be of almost nothing. You’re looking at the back of someone’s head for five minutes.” The minimum recommended processor for a four-camera network starts with a 2 GHz Pentium 4. Stepping up to a 25-camera network leaps you into at least a dual 3 GHz Xeon server. Obviously, these specs predate the shift from NetBurst to Core CPU microarchitecture, but you get the point. Axis’ Nilsson comments that a modern laptop can only handle about 10 cameras, and a hefty server starts to buckle past 100.
Neal Jensen, president of Radius Security, advises resellers not to try dropping an NVR application on an existing server. The processing load is potentially too great, especially if that customer keeps adding cameras. Instead, deploy it on a new, dedicated server preconfigured with the NVR application and lots of storage—as in terabytes. Even heavily compressed video will still plow through terabytes at megapixel resolutions. This is doubly true in businesses where companies are required to maintain recorded data for at least 30 days. Jensen points to a recent 16-camera addition Radius did at a car dealership; the installation called for adding another 6TB to 8TB of storage capacity. If this level of storage can fit in a tower, fine, but odds are you’re now looking at some kind of rackmount NAS opportunity. Also, don’t make the mistake of thinking that one server may be enough. “On a monitoring station,” notes Axis’ Nilsson, “if you’re looking at a $3,000 Nortel or Cisco switch that has a 10 Gig uplink on the backside to the backbone, you can technically run full frame rate video for 4,000 cameras in that 48-port switch. So the network is not the problem. The problem is in monitoring. Once you want to start monitoring more than, say, 30 cameras, you need to start installing multiple servers with very good graphics cards and monitors that can deal with the video decoding. You see this in places like traffic monitoring centers and casinos.” There are plenty of ways to lighten the bandwidth load in various places, and the first is not to oversell your IP technology. Just because you can do 60 fps at 1.3 megapixels doesn’t make it a good idea. The customer will be dazzled until he sees that 70% of the solution cost is in storage for that monstrous data load. Truth be told, this is surveillance, not cinema, and most people don’t see a difference between 15 and 30 fps. Perhaps you can find software flexible enough to only record at 2 fps unless there’s an alarm trigger, at which point the NVR can shift into 15 or 30 fps. Does your client actually need megapixel resolution, or can he get by with CIF if there is no alarm in play? With just a few tweaks like these, you can wipe away 80% to 90% of the storage load. And there’s no rule that says you have to centralize everything. If the customer doesn’t have high bandwidth connections between multiple localities, don’t make centralization of the surveillance storage your primary goal. Consider taking the 40 or 100 or whatever number of cameras at each location and feeding them into one or two servers at each branch. These servers can spool into local NAS storage and use management software able to track across multiple locations. Or if you want to get really slick, set up a SAN to tie these multiple storage pools into one centrally managed array. (Refer back to our storage cover stories in RAM issues 65 and 66 for more on this topic.) There’s no concrete need to keep streaming all of this data back to headquarters. ...more |
||||||||||||||
 Back to top
|
||||||||||||||
Copyright © 2007 RAM Magazine. All rights reserved.
Do not duplicate or redistribute in any form. |
||||||||||||||
