Eyes on Efficiency: DSS Systems Seek Performance at the Right Price




The next generation of Digital Security Surveillance systems is looking to innovative SDK approaches as developers seek to expand DSS performance without a higher TCO price tag.

Intelligent Video Analytics (IVA) is the technology that automatically identifies and tracks objects from video streams and analyzes motion. Coordinating with the database to extract video intelligence pertinent to specific purposes, IVA can increase the efficiency and effectiveness of video surveillance systems. MarketsandMarkets predicted in its Worldwide Market Forecasts & Analysis (2012-2017) that IVA would be the fastest growing technology segment of the surveillance (IT) sector in the next three to five years. The report expects the IVA market to reach $867.8 million by 2017, up from $180 million in 2011, with an increasing CAGR of 30.4% from 2012 to 2017. And IVA is driving surveillance market growth, which IHS Technology predicts will grow by more than 12 percent this year (2014) from $14.1 billion in 2013 to $15.9 billion. [1]

Now, video surveillance no longer stops at recording and playback. Intelligent Digital Security Surveillance (DSS), based on IP networking, captures, processes, analyzes and transmits video data to a control center in real time, with all cameras and other field devices connected to a network platform associated, in turn, with databases. Controllers at the field end demand the computing power to fuel multiple analytic and statistical tasks.
fig1
Figure 1. Analysts predict double-digit growth for the global video surveillance market.

Price Competitiveness on the Rise
The multitude of distributed cameras, embedded systems and other peripherals that DSS systems can require make price a factor when building a DSS system. Though price is just one issue, it is often a top priority—at least to everyone other than the system architect. While ARM-based systems generally offer lower power consumption and overall cost, they fall short of the computing demands required for large-scale video surveillance applications. And while x86 based systems provide the necessary computing power, they also generally come with a heftier price tag. However, with the introduction of software-based solutions to improve video transcoding efficiency and free up CPU resources, reduced hardware requirements in x86 systems have helped to lower overall TCO and make them much more price competitive.

In addition to real-time analytics, current user demand in digital surveillance is for higher definition video output in uncompressed form. HD quality is no longer sufficient to meet market standards, as surveillance operations benefit greatly from clearer, more information-rich video images. Uncompressed full HD content at 1080p is becoming the norm, demanding significantly increased computing power.

To enable real-time analytics, remote management over the network is needed. At the hardware level, this means remotely monitoring and controlling the status of all devices in real time from a back-end server, with system administrators executing basic trouble shooting over the network to prevent system shutdowns and the associated cost and manpower expenditures.

At the application level, remote management allows field-end controllers to process and analyze the data captured by field-end devices, then forward the information to the control center in real time, making the control center’s immediate response possible. Further analysis of the gathered data from historical, geographical or statistical perspectives can yield information to support policy-making and other high-level tasks.

DSS field-end embedded devices deployed in outdoor or semi-outdoor environments require robust and ruggedized construction, equal to the harsh conditions of use. To succeed in such 24/7 mission-critical environments, a reliable, secure infrastructure is crucial, providing efficient management while circumventing the liabilities of unplanned downtime. A zero-downtime platform such as this is likely to comprise imperviousness to moisture and contaminants, resistance to shock and vibration, wide operating temperature range and surge protection.

Limitations of Conventional ARM-based and x86 Systems
For large surveillance systems, such as those utilized for purposes of monitoring urban activity or road/traffic, field deployment is costly due to the need for large numbers of cameras, embedded systems and other peripherals. For years, ARM-based controllers have been widely used in field deployments of large-scale video surveillance applications due to their simplified CPU architecture and lower power consumption and pricing—compared with x86 systems.

However, with limited computing power, an ARM-based controller typically focuses on a task such as “record and playback video data.” When the application calls for more increased functionality—to detect specific occurrences such as traffic infractions, for example—the number of controllers and cameras required increases.

fig2
Figure 2. The controller and camera count rises when DSS systems are called upon to identify traffic infractions and other specific occurrences of interest to law enforcement.

Though ARM-based solutions are making great progress in computing and graphic performance, they still fall short of demands from more advanced applications such as IVA. ARM-based architecture also limits flexibility in system design due to hardware/OS/software compatibility issues. For instance, when hardware or OS is upgraded, all application programs must be rewritten—a costly and inconvenient prospect for system integrators.

In comparison, x86 architecture-based systems excel in computing power, plentiful I/O and multitasking capabilities, as well as offering richer software resources and backward compatibility. Even these advantages, however, have not been able to offset price issues in large-scale field deployment. The emergence of Intel® Media Software Development Kit (Intel® MSDK) technology though, has begun to turn things around.

Surveying DSS Choices
Every security surveillance system requires considerable capital investment. When the market is experiencing a paradigm shift, decision-makers involved in surveillance system deployment will have to consider whether the system they are choosing will cater to their future needs. Other considerations include the ease and costs of upgrading and expanding.

Unlike computing platforms with the CPU resources to support multitasking, single-purpose digital security surveillance systems with limited CPU resources, while less costly, may lack the full-spectrum functionality next generation surveillance systems need.

Another consideration is the interconnection of all devices necessary in intelligent systems. An embedded system with diversified I/O capability can offer the reliability and convenience of the inter-device connection the system demands. This could include, for example, lighting control through RS-232 ports, Wi-Fi connection to a backup server and sufficient GbE ports for multiple IP cameras.

As the costs of processors, storage and Internet connection decrease, and new technologies quickly increase single system channel capability and quality, high definition surveillance systems are rapidly becoming more easily realized. The number of channels, image resolution and signal quality and additional functionality, such as real-time image monitoring and analytics for various circumstances, dictate the maximum system scalability possible to ease the burdens of system upgrade.

In choosing computer systems, DSS system integrators should consider software compatibility issues between different OS versions and programming languages to ensure higher flexibility in software use. Backward and forward compatibility is important for conserving system development costs.

Present as hidden expenditures of which customers may be unaware, maintenance costs form one component of TCO. The number of embedded device sites in a DSS system must normally exceed a hundred. With such numbers, if maintenance and management require too much on-site attention, costs quickly add up. Using easily executed remote control can reduce or eliminate the need for individual attention to system status.

In addition, a reliable system should be able to withstand harsh environmental conditions such as extreme temperatures, surge impossibility in cabling and vibration and shock, to further conserve maintenance costs.

Innovation Expands SDK Benefits
x86 systems, notable for high computing power, rich I/O and abundant software resources, present a favorable choice for developing next-generation intelligent DSS solutions, and the emergence of Intel MSDK technology continues to make x86 systems more price-competitive.

The Intel MSDK provides the fastest performance possible for Intel® processors equipped with Intel® Quick Sync Video. Equipped with a driver to offload transcoding tasks (decoding, processing and encoding) from CPU to GPU, the Intel MSDK increases speeds and reduces CPU loading. When the GPU performs video transcoding tasks, the bulk of CPU resources become available for other operations such as data and peripheral control, improving overall computing response and performance.

Intel’s partners go one step further by leveraging the benefits of the Intel MSDK and contributing homegrown innovation to further enrich the technology. For instance, the ADLINK MSDK+ capitalizes on Intel® MSDK technology. As shown in Figure 3, the ADLINK MSDK+ can handle operations beyond those the Intel® Media SDK supports, including mux/demux of media container files and real-time transport (RTP) receiving and streaming.

fig3
Figure 3. Lessening the CPU’s transcoding burden by shifting some tasks to the GPU frees CPU resources.

The ADLINK MSDK+ can assist in demuxing video and audio elements from a container file, allowing the Intel® MSDK engine to focus on processing the video elementary stream, extracting video data pertinent to application purposes and then remuxing extracted video with audio into a container file.

RTP facilities expedite the streaming of video and audio elements over the network by not focusing on overhead information.

Embedded systems with ADLINK MSDK+ support enhanced graphic performance, boosting media streaming and reducing CPU loading. As shown in Table 1, if a 10 second 1080p video is to be transcoded to 480p via CPU, only two transmissions can be managed simultaneously before CPU resources are used up. In a system featuring ADLINK MSDK+ support, as shown in Table 2, however, the system utilizes GPU to offload transcoding tasks from the CPU, transcoding a 10 second 1080p video to 480p with more than 12 flows supported simultaneously, with only 11% of the CPU resources occupied. This represents a significant reduction, up to 80%, of CPU loading, with an impressive 250% increase in transcoding speed.

table1
Table 1. Transcoding 1080p to 480p video via CPU.

table2
Table 2. The results of using ADLINK MSDK+ support to offload transcoding chores from the CPU to the GPU.

When applied to video surveillance, the liberated CPU resources can control more peripherals, providing analytic services to more video channels. For example, in the past, each x86 computer could support only two digital cameras because dealing with graphics processing tasks occupied most CPU resources. Currently, however, an embedded x86 computer featuring ADLINK MSDK+ can offer connection of up to 12 cameras (depending on network bandwidth), with even faster video streaming at each channel.

Conclusion
Facing the rising demands of the video surveillance market for more intelligent features and functions, system developers and integrators must discern how to balance performance and cost, meeting customers’ needs while satisfying their own budgetary concerns. While there are many other considerations, such as ruggedization, connectivity, scalability and the other issues previously outlined, TCO is key.

Third-party vendors seeking to increase video surveillance system performance while significantly reducing CPU use are bringing added benefits to the Intel® MSDK. Offloading transcoding tasks from the CPU to the GPU frees the processor for other tasks. With the lowering of total costs for x86-based field deployment, a competitive alternative to conventional approaches becomes available.


Zane Tsai is product manager, I/O Platform Product Center, ADLINK Technology.

Share and Enjoy:
  • Digg
  • Sphinn
  • del.icio.us
  • Facebook
  • Mixx
  • Google
  • TwitThis

Tags: ,

Extension Media websites place cookies on your device to give you the best user experience. By using our websites, you agree to placement of these cookies and to our Privacy Policy. Please click here to accept.