Smart video solutions are now becoming a reality, instead of hype. Anders Laurin, Executive Vice President, Axis Communications, discussed real-life analytics for real-time risk assessment at the Global Digital Surveillance Forum, part of SecuTech Expo 2007, in Taipei, Taiwan.
What a difference two years can make. Back in 2005, the hype around intelligent video raised expectations beyond what was possible. It did not work because it did not reflect realityit was a wish list that reflected a future direction for network video.
Two years on, it is no longer hype about the future. Axis Communications is putting the smarts into network video today. Just as intelligent video uses analysis to filter out irrelevant data, we have filtered through the wish list and begun delivering realistic intelligent network video applications that are delivering real benefits right now.
The approach has changed but the challenge still remains. We still need to solve the same problem: We have information overflow and we need tools to assist operators. We will need them for the bigger systems we will be building in the future.
Analog systems are generally around 10 to 50 cameras. Fifty cameras was considered large for an analog system. When we talk about IP-based systems, we are often talking about thousands of cameras. Axis Communications has been involved in several installations comprising 10,000 network cameras. Installations with 1,000 network cameras are common now.
The Intelligent Video Market: A work in progress
The intelligent video market today is a work in progress. It is still early and the market under development. Intelligent video remains as one of the key development area for video surveillance, with most manufacturers investing in this area. But the difference now compared to a few years ago is that the industry is taking a more realistic approach today.
Most companies have realized it is not the complex algorithms they should try to sellit is the more realistic ones. The focus is now on realistic intelligent video functions like motion detection, tampering alarm, object counting, object direction and object tracking which is becoming increasingly popular.
The intelligent video market is moving in four directions:
Embedded: Redundant CPU power
The first is an embedded solution where products use redundant CPU power in the camera. The same CPU is used for networking or the camera functionality. That means there is no need for special intelligent video productsthey use the standard, mainstream products- -with no extra hardware cost.
There are drawbacks with an embedded solution. The algorithms need to be smarter and smaller, as memory is shared with the camera. Users also have less CPU power because all the other activities, like camera processing, are on the camera too.
Co-processor: Extra CPU power
The second solution is the same as the embedded solution except that it uses a co-processor. This has the advantage of providing unlimited CPU power and memory, so huge systems can be built with this approach, but it comes at a price. The co-processor solution costs significantly more and requires high power consumption.
The most critical part of this solution is heat generation that makes it unsuitable for the camera, as many manufacturers discovered when developing these types of network camera products. That makes this solution more suited to a video encoder, rather than camera-based products.
DVR or dedic ated PC
The third solution is perhaps the most common todayapplications running on DVR or an external PC. This gives you unlimited CPU power on the PC and unlimited memory. There are modules available for this type of solution and it has the benefit of using a known PC architecture.
While that is perfectly fine, there are some negatives with this approach, including limited CPU power in the DVR and the need for a dedicated PC. That makes this slightly more expensive as a total solution.
Using a DVR or dedicated PC might be the best solution today for small systems, but it is not the answer for large ones.
Utilize both embedded power and an external PC
The fourth approach is a mix of embedded power and an external PC. It delivers numerous benefits over the other three alternatives. It provides unlimited CPU power from the PC and unlimited memory. It is scalable and requires less bandwidth. It can be complex to program, but that will improve over time.
The key to the power of this solution is using metadata and sending only essential information. For license plate recognition, for example, you would only need to get and send the information you needthe character string, the time and date. You do not need to send the information at 30 frames per second resolution.
This hybrid approachcombining an embedded solution with a smart architecture that supports intelligent video in different places in the systemis most likely the way to go for intelligent video in the future.
Distributed is a key word for the future. The more intelligence you can place closer to the lens, the better. A good network camera should act as an intelligent gatekeeper. There is already a significant amount of intelligence available in network cameras today, including alarm handling, motion detection, audio detection, and in some cases, event managers.
Video intelligence requires a large amount of processing power. In the future, we will see greater functionality, more advanced intelligence and more intelligence on-site. These improvements will be delivered hand in hand with more CPU power in the camera.
Active Tampering Alarm: Putting intelligence in the network camera
Demanding and hostile environments can put extreme demands on video surveillance solutions. This is especially true for those operating in industries like transportation, industrial or resources and education, as well as government monitoring of public areas or prisons.
How do you monitor a large system of hundreds or thousands of cameras to make sure they a re working properly in tough environments where weather, accidents and acts of violence or vandalism can cause them to be redirected, blocked, covered or sprayed? You do it by including intelligence in the network camera.
Axis has done that through the development of a new intelligent video function called the Active Tampering Alarm.
The Active Tampering Alarm enables you to detect redirections, blocking, covering and spraying of the camera quickly and easily.
A camera could be redirected accidentally by cleaning staff or a strong wind. For example, the camera誷 view could be blocked by a parked vehicle or equipment placed in front of it, or it could have been spray painted or covered in a deliberate act of vandalism. The operator is alerted immediately about any disruption or attempt to disrupt normal camera operation, so they can take action immediately.
The Active Tampering Alarm is an example of a new realistic intelligent video function delivering the benefits of distributed intelligence. Axis has placed video analytics in the camera with no central processing required, as it is all done in the camera. It scales effortlessly as well, because you do not need more CPU power.
A standard Application Programming Interface (API) from the camera makes it simple for our Application Development Partner (ADP) Program to integrate support for Active Tampering Alarm in their applications.
The Active Tampering Alarm function is included free of charge in a number of Axis products designed for tough environments.
Real Applications, Real Results: Stockholm Transport Security Project
An intelligent network video surveillance system using Axis Communications technology has played a key role in the success of a real-life application for Stockholm Transport.
Stockholm Transport is responsible for 650,000 public transport passengers every weekday on buses, commuter trains and local railways. In 2005, it decided to embark on a security project aimed at increasing passenger and employee safety. The first security system launched under the project was an IP-based alarm and video system commissioned at major train stations at the end of 2006. Cameras covered all areas at major train stations, delivering an overview of all facilities and buildings, as well as providing the image quality needed to identify individuals.
Stockholm Transport decided very early on to use network cameras and to include some intelligence in the system to address issues, such as unauthorized track access by people taking shortcuts across tracks between platforms. An alarm is automatically raised in the system as soon as unauthorized track access is detected. The first installation was extremely successful and resulted in a rapid rollout to include all stations in the Stockholm Metro system. Open system-based fire alarms in tunnels and train stations have been integrated in the video and alarm management system. In the event of a fire, an alarm is sent to the operator and a camera can be viewed immediately.
Another trial project saw cameras in five buses integrated into the video management system for true video surveillance with functionality. It included an on-board recording when the bus was in service, along with system failure monitoring as an intelligent function for camera tampering. When an alarm was tripped, the operator received a live view, including location information. Because the system was integrated, they can see the location of any guard near the site of the incident and call them to investigate if needed. The success of this project resulted in a decision to equip all buses with four to six camerasa total of 10,000 cameras in 2007all ultimately connected wirelessly to the project central safety center.
This is the future of intelligent network video, in action today. There are many more potential applications and opportunities. While there are still not many real installations yet, the tide has turned. The focus has shifted from hype about potential applications to the delivery of realistic ones.
We are building a market from the bottom up now, not the top down, as intelligent video takes its natural place as part of the network video surveillance market offering realistic applicationsand real benefits.