Lights, Camera, Action:Tuning Megapixel Cameras for Real Life
Editor / Provider: a&s International | Updated: 4/13/2011 | Article type: Tech Corner
Megapixel cameras see great amounts of detail, but require a carefully selected supporting cast of storage, networks and displays for an outstanding performance. a&s finds out from experts how to set up video systems for real-life megapixel surveillance, as well as how to specify cameras.
Most megapixel demonstrations take place outdoors on bright cloudless days, giving endless vistas of crystalclear detail. While some indoor installations also require a high level of detail, most spaces do not require the wider range of view. “Indoor megapixel surveillance is less impressive because of the close distance,” said Patrick Lim, Director of Sales and Marketing for Ademco Far East. “Outdoor, wide-area surveillance sees the most distinctive requests and appeal for megapixel.”
Sensitivity levels on the camera's spec sheet need to be taken with a grain of salt, as they still require added lighting. “When it comes to comparing camera sensitivity at low-lux lighting levels, be careful, they usually refer to the light at the sensor, not the ambient lighting in any given scene,” said Mark Wilson, VP of Marketing, Infinova. “To get these lux levels at the sensor, the ambient lighting at the scene may have to be many times more.”
Megapixel surveillance expansion is relatively simple, as new devices are generally configured by the VMS. “With an IP system, you can add or move a camera by simply mounting it and plugging in a single structured cable to deliver all video, bidirectional data and PoE,” said Becky Zhou, APAC Sales Director for Arecont Vision. “In most cases, an IP address is automatically assigned upon detection of the new network edge device and you're in business.”
Large-scale sites usually have network infrastructure in place to support megapixel-camera expansion. “You can scale a system easily if all the information travels over the same Ethernet cables,” said Ian Johnston, CTO of IQinVision. “With an IP-centric design, security video, VOIP telephone calls, physical access control systems as well as their conventional LAN data traffic are all combined on the same network. This enables large enterprises to focus their dollars on one department — IT — and gain huge benefits across the board.”
Scalability depends on how many devices are on the network. A small system simply requires an NVR and basic switch, with bandwidth generally not being a problem, but gets more complex with more cameras. “Once you're over 100-plus cameras, you have to think carefully about how to manage network bandwidth,” said Andrew Pigram, Technical Director at Norbain. “Systems of this size will probably demand multiple places to view and record. You need to think about the camera, multicast and what kinds of switch protocol to minimize bandwidth and get information sent to the right location.”
Compression has a direct effect on bandwidth and recording, as not all recording platforms support H.264 or MPEG-4 compression for third-party megapixel cameras. “For a remote island facility, if you remotely connect to the site, on some platforms not all codecs are available to a remote user,” said Stephen Moody, Security Development Manager for ViS Security Solutions. “Bandwidth implications come into play. You would have to start prioritizing the network to make sure you're getting the proper flow of traffic. Using quality of service applications often rectifies this and provides you with the correct bandwidth suitable for the application, even if the codec is not remotely supported.”
[NextPage]Ideal traffic flow on the network means the health of the network cameras does not affect other online services, such as teleconferencing, digital signage or television content for large organizations. “If there's not enough bandwidth and the camera uses H.264, the image distorts and you miss a lot,” Moody said. “But with M-JPEG, you have a clean flow of image. H.264 or MPEG-4 will flow, but if it's not set up correctly over the network, the image tends to pixelate, or you lose the image and get image distortion.”
The biggest installation variable is bandwidth capacity available or intended, as well as the distances of cable runs, Zhou said.
Bandwidth varies between WAN and LAN, with many megapixel cameras supporting triple streams to push live, low-resolution video to phones or home computers. “On a local network, you can see whatever you want to see, including megapixel resolution,” said Todd Pinnell, PM for Video at Speco Technologies. “But there's a limit on what you can put on that network.”
Cropping megapixel images may reduce bandwidth requirements. “Since you can crop a network camera with an arbitrary aspect ratio, you can deliberately set up the camera not to record the sky or the immediate foreground — areas you most likely don't care about,” Johnston said.
The more cameras a site has, the more bandwidth and storage is needed. “We always advise the customers to invest wisely by utilizing event or intelligent video triggering to reduce unnecessary recording,” Lim said. “This also reduces the amount of information the customer has to store and review.”
Fortunately, storage costs are dropping, making megapixel cameras less prohibitively expensive. “A 3-terabyte hard drive retails for only US$179 now,” Johnston said. “That's practically free in today's terms, and two to three years from now, 3 terabytes will be laughably inexpensive and small.”
Redundancy in storage is critical for some sites, but is not supported on all VMS. “You need to look at the software, then buy the right PC servers and storage hardware to match,” Pigram said.
[NextPage]Megapixel on Display
Megapixel images are not always in the right aspect ratios, making them tricky to display. “We've come across applications with people deploying 10-megapixel cameras but don't have the monitor or workstation with the capability to display 10-megapixel images, which defeats the purpose,” Moody said. “If you don't have the right monitor or right workstation for a 10-megapixel camera, you get motion ghosting.”
Correctly displaying megapixel video depends on the software used. Some VMS allows users to set the aspect ratio by camera type. “The display connects directly to the PC running the VMS and is therefore media-independent,” said Karen McCarrison, PM at IndigoVision. “Cabling is determined by the PC graphics card and monitor interfaces. When viewing HD video (16:9) on a 4:3 format display, the aspect ratio of the video is maintained via the software. The same is true when viewing standard definition (SD, 4:3) on a wide aspect ratio monitor.”
Video walls are more challenging for large projects, due to the greater number of camera inputs. “Sometimes, megapixel cameras don't show up in the settings or show up in a weird aspect ratio,” Lim said. “Because different video walls have different sizes, video wall makers do not usually work extensively with security systems.”
While analog cameras just needed to be plugged in to display, pushing network megapixel video to the screen is more complex. “A project we worked on before had the video wall trying to pull megapixel video, but it came out distorted and did not display,” Lim said “IP streams are easier to pull in, but configuring software is not so easy. The video wall may not have the decoder for that stream.”
Security VMS packages will need to support the various resolutions available for megapixel cameras and display appropriately. It may maintain the aspect ratio based on the camera signal or display black bars where necessary, McCarrison said. Another option is to stretch or squeeze the image in the VMS, which loses the image's aspect ratio.
Integration There are many features on megapixel cameras, but few actually work with security management systems on converged platforms. “It's quite misleading with alarm I/Os or audio; you think you can connect to speakers or integrate perimeter intrusion detection systems with the camera I/O, but integrating with the VMS is a challenge,” Moody said. Not all features of the camera are supported, particularly in third-party camera applications.
[NextPage]ONVIF and other standards are good in the long term for inoperability, but systems have yet to be 100-percent open. “ONVIF at the moment is limited to live viewing directly from the camera,” Pigram said. “Replay and retrieving video from where it's recorded will be addressed in the future.”
“The end user or customer expects an ONVIF camera to work directly, but they will be disappointed,” said Gerrit Schreiber, Senior PM at Basler. “You will see in many tenders that ONVIF is a must, but they do not use ONVIF at all.”
Although ONVIF is not fully compatible between versions 1.01 and 1.02, particularly for PTZ service, it depends on each member to state which version they support. “ONVIF is not specifying how the video needs to be encoded,” said Sachin Khanna, PM for CCTV, Bosch Security Systems. “ONVIF specifies the control protocol on how to setup and get a video stream from a network camera or an IP encoder.”
Most new video devices have basic support for ONVIF, PSIA or both. “In the coming years, a winner will naturally emerge, but in the interim, having two specs to choose from is much preferred to having 50,” Johnston said.
The detail megapixel cameras offer lends them to a wide range of applications. “Megapixel cameras provide crisp, detailed images that enable operators to read fine details in the distance like license plates, to see the face of the driver inside an automobile, to read the suit on a playing card or to see the denomi-nation of money used in PoS and gaming applications,” Zhou said. “These traits lend themselves to a myriad of applications for both security-centric and for business and process management applications in areas such as retail to monitor purchasing habits, quality control and automated manufacturing, medical for surgical applications and so on.”
Just as smart phones have become more open for apps, network cameras are opening up to new users. “We're opening up the camera to our partners to be able to put their application code directly in the camera,” Johnston said. “For the moment, the most popular ideas have been integrated analytics and local recording for network fail-over systems. As the system matures and an application marketplace evolves for the security industry, many exciting possibilities will soon appear on the horizon.”
As megapixel cameras have b e c ome mo re c ommon, the y are being specified for projects more frequently. However, some expectations are beyond what the technology can offer today, presenting a gap between reality and fantasy. “We've seen an actual spec written, asking for 3-megapixel cameras recording 25 fps at the best quality and stored to a 500-gigabyte hard drive,” Lim said. “The storage was expected to archive for three months. If we have 20 units of 3-megapixel cameras recording at 25 fps at the highest resolution, it will take half an hour, maximum, to use up 500 gigabytes. The reality is many, even consultants, don't understand how big 3-megapixel is at 25 fps.”
From a solution standpoint, adapting megapixel cameras includes upgrading back-end displays and storage, said Alf Chang, Senior Consultant for A&S magazines and a former installer. If deploying megapixel cameras requires a complete computing upgrade, the migration to IP faces many roadblocks.
[NextPage]A common complaint about system specifications is people sometimes cut and paste words from A&E and other product documents. “It produces not a true system, which could never work as described or is very inefficient,” Pigram said. “The normal cut-and-paste job is not truly understanding what customers want and building a system like a bad jigsaw puzzle.”
The bottom line is to not only look at data sheets but figure out if the camera fits user requirements, especially when it comes to motion, low-light handling and stable frame rate without frame drops, Christensson said.
Some users assume 10-megapixel is better than 1 or 2, but do not see overall problems with lenses and storage. “They want real-time and 30 fps — these expectations cannot be served,” Schreiber said. “If you have a well-educated integrator, he will know what is possible and communicate with the end user. But there are still some very ridiculous requirements, such as seeing with no light or looking behind walls.”
While megapixel sees a great deal, it still has limitations. “From Hollywood, users see ‘CSI' or ‘Mission Impossible' closing in on villains,” said Vance Kozik, PM at StarDot Technologies. “However, a camera can't see 500 meters to read a license plate. Hollywood has presented something that can't be done yet.”
While a 5-megapixel camera equals the resolution of 15 D1 cameras, it doesn't mean users can digitally zoom half a mile away and recognize a face, Kozik added.
As Hollywood investigators identify villains from reflections or at great distances, user expectations have increased for megapixel. “This market is close to the consumer one,” Schreiber said. “Most are not highly trained or in engineering. When they look at TV and watch ‘CSI Miami,' they will see the police identifying a car from 1 kilometer away. This is fake, but the fakery of TV and the movies has moved to their industry expectations.”
While a megapixel camera might not spot a terrorist three miles away, increases in processing and better illumination have improved image capture. Technology is closing the gap between reality and fantasy, with faster processors ensuring more accurate and better analytics as well as sharper images. High-resolution camera installation requires planning for displays, storage and networks, as well as overcoming integration hurdles. Extra preparation and understanding will go a long way toward ensuring solid megapixel camera performance.