It’s possible that the most expensive and ineffective video surveillance equipment in any organization is the security staff assigned to watch monitors. Video surveillance literature often cites is a study by the U.S. National Institute of Justice that found: “After only 20 minutes of watching and evaluating monitor screens, the attention of most individuals has degenerated to well below acceptable levels. Monitoring video screens is both boring and mesmerizing.”
The industry’s answer to this problem? Video analytics solutions that automate most of the “watching.” In this post, I want to look into the performance considerations for such solutions deployed in edge devices and the role that 3rd generation Intel® Core processors can play.
Video analytics solutions use complex algorithms to determine changes, such as an object moving, within a prescribed view. Video analytics also provides a way to search recorded video for specific events. This is a real time-saver compared to having someone watch fast forward through hours of video from multiple cameras.
In digital security surveillance (DSS) systems, video analytics software is generally set up to send alerts to a centralized control room, as well as mobile devices used by on-premise security. Staff can view the video and determine a proper response.
A typical IP-based video surveillance system uses multiple video cameras, some of which may be analog and some digital IP network cameras (see Figure 1). The analog cameras feed into encoders which digitize their video. The digitized video is then transmitted to a central control room, where a server processes the video for display and storage.
Figure 1. Diagram of a hybrid (analog and IP camera) DSS system.
As the diagram shows, video analytics solutions are often deployed on standard off-the-shelf servers at the edge or in a central location. Video analytics can also be run on NVRs or embedded in video surveillance devices such as network cameras and encoders. Many DSS solutions use both.
Centralized server-based solutions generally collect full streams of video for analysis from all across the network. For more on these solutions, see my earlier post “High Performance, High Bandwidth for Large DSS Systems.” In edge server solutions, the video is analyzed locally and can be briefly stored and then deleted after a set period of time if no event was detected. This is an excellent strategy for reducing the bandwidth demands of a DSS system.
A more recent strategy for edge video analytics is to divide the work between the camera and video encoder hardware. The camera performs simple analytics while the encoder performs more sophisticated analytics. This means that the encoder has to have a processor powerful enough to do more than just digitize and compress the video signal. It has to be powerful enough to perform video content analysis. The advantage of this approach is that it eliminates a device—the edge server—but still makes optimal use locally of high resolution video, while minimizing network video traffic.
Some DSS architectures are beginning to use this approach and more will surely follow. It’s a particularly good strategy in systems using encoders to stretch existing investments in legacy analog cameras.
To enable video content analysis in an encoder or a sophisticated camera, the device needs to become an intelligent system. In fact, the more sophisticated the video analytics in terms of combining mathematical, statistical, signal and image-processing techniques with machine learning, pattern recognition and other types of algorithms, the more intelligent the device has to be. When you add the tremendous amount of data that high resolution video cameras generate to this equation, the processing requirements are intensive.
On most DSS devices that handle video, a majority of the processing capacity goes to system control and image processing tasks. This leaves limited capacity for video analytics. Adding a dedicated analytics processor is a possibility, but increases the bill of materials (BOM) and the price of the ultimate product. IMS Research sees a different way. In its report on video trends for 2012, IMS predicts that not only will the types of applications that can be performed at the edge increase, but that this trend will be driven by the availability of more powerful general purpose processors and the refinement of video analytics applications that will make them less processor-intensive.
There’s little reason to wait though if you’re a designer of DSS systems. The 3rd generation Intel Core processors offer all the required performance for today’s video analytics solutions. What’s more, the broad selection of off-the-shelf boards from members of the Intel® Intelligent Systems Alliance makes it easy to meet the small form factor needs of cameras and encoders, plus maintain competitive pricing.
A good example is the Portwell PCOM-B219VG, a Type 6 COM Express Compact (95mm x 95mm) module based on the 3rd generation Intel® Core™ processor and Mobile Intel® QM77 Express chipset (see Figure 2). With multiple cores (from two to eight), as well as Intel® Hyper-Threading Technology, this processor family provides excellent multithreaded processing and multi-tasking. It also includes Intel® Turbo Boost Technology 2.0 to boost performance for specific workloads by increasing processing frequency. Equipped with a low thermal design power (TDP) processor (as low as 17 W), the PCOM-B219VG can provide superior performance while sipping power in harsh environments from as low as -40 ℃ and up to 80 ℃. Support for one PCI Express*x16 Gen 3 lane (8.0GT/s) and USB 3.0 more than double I/O throughput over the previous generation. Seven PCI Express x1 Gen2 (5.0 GT/s) are also included. Support for up to 16GB ECC DDR3 1333/1600 MT/s SDRAM on two 204-pin SODIMM sockets provides plenty of memory headroom.
Figure 2. Portwell PCOM-B219VG Type 6 COM Express Compact (95mm x 95mm) module.
Of particular note for DSS applications like encoders is the upgraded graphics engine of the 3rd generation Intel Core processors. The Intel® Core™ i7-3770 processor, for example, can encode four channels of 1080p 30 frames per second (fps) video and decode 20 channels of 1080p 30 fps video. Though admittedly a more powerful processor than available with the PCOM-B219VG, it provides an idea of the power of these processors.
For maximum performance, media processing applications can be optimized using the Intel® Media Software Development Kit (Intel® Media SDK 2012). This cross-platform application programming interface (API) for developing media applications supports the hardware-accelerated video encoding, decoding and transcoding provided by the Intel® HD Graphics 4000 integrated into 3rd generation Intel® Core™ processors. The Intel® Media SDK 2012 significantly reduces the time for coding media applications that take can advantage of this underlying hardware performance. The Intel® Media SDK 2012 supports a range of codecs and features, including H.264, MPEG-2 and MVC video encoders and decoders and various video processing filters. The key advantage here is that fixed-function hardware blocks perform the codec operations, minimizing processor utilization. This, in turn, frees up additional headroom for video analytics and other tasks.
In addition to the performance boost of Intel HD Graphics 4000, 3rd generation Intel Core processors have one more thing up their sleeve: Intel® Advanced Vector Extensions (Intel® AVX). This set of instructions for doing Single Instruction Multiple Data (SIMD) operations on Intel® architecture supports the high-performance 256-bit vector and matrix processing crucial to applications like video analytics.
Of course, one of the big advantages of COM Express modules like this one from Portwell is that it enables designers to partition host processors from proprietary baseboards, thereby minimizing current and future design risks during the initial phase of development. Separating the CPU-upgradable module from system specific I/O carrier boards safeguards development investments and lowers total cost of ownership. In addition, companies like Portwell can also provide services to clients on the carrier board design and development, review schematics and BIOS customization.
Have some thoughts on moving more video analytics to the edge with intelligent encoders and cameras? Please send them my way. In the meantime, I recommend checking out the white paper: Intel® Digital Security Surveillance System Media Performance Benchmark Methodology.
To learn more about DSS solutions based on Intel processors, see our top picks on the subject..
Portwell is a Premier member of the Intel® Intelligent Systems Alliance.
Roving Reporter (Intel Contractor), Intel® Intelligent Systems Alliance
Associate Editor, Embedded Innovator magazine