The IEEE ComSoc SCV chapter April 14th, 2010 meeting was very well attended with more than 80 people present. This was our first joint meeting with TiE- The Indus Entrepreneurs organization. The meeting was titled, “Architectures and Applications of Video Surveillance and Video Analytics Systems.” and featured talks plus a panel discussion on those topics
The speakers scheduled to participate in the talks and panel session were Professor Suhas Patil, Chairman and CEO of Cradle Technologies, Basant Khaitan, Co-founder and CEO of Videonetics and Robb Henshaw, VP Marketing & Channels, Proxim Wireless. Robb Henshaw who was scheduled to speak on “A Primer on Wireless Network Architectures and Applications for Video Surveillance” could not attend the meeting due to illness and was replaced for the presentations section of the evening by Alan J Weissberger, IEEE ComSoc SCV chairman. The panel session was moderated by Lu Chang, Vice-Chairman of IEEE ComSoc SCV. This article has been co-authored with Alan J. Weissberger, who contributed the comment and analysis section, raised several unaddressed but nevertheless pertinent questions and also provided references to background articles on video surveillance and video analytics.
While presenting on behalf of Robb Henshaw on Wireless Network Architectures, Alan J. Weissberger noted that several new technologies are now converging which will make video surveillance a growing market and viable business. These include: higher-quality IP digital cameras, improved and cost-effective video compression technologies (e.g. H.264/MPEG4 and HDTV), fixed broadband point-to-point and point-to-multipoint networks (including fixed WiMAX and proprietary technologies), and mobile broadband (including 3G+, mobile WiMAX and LTE).
To support the claim of a growing market for video surveillance and video analytics, Alan cited several key examples of applications for these technologies such as: security and surveillance applications, emergency and disaster management, asset and community protection by monitoring of buildings and parking lots, public entry/exits, sensitive areas such as ATMs, as well as high-traffic areas like highways, bridges, tunnels, public areas such as parks and walkways, infrastructure like dams and canals and buildings like a cafeteria, halls and libraries. Other applications include securing of sensitive areas like runways and waterways, perimeter security for military installations, remote monitoring of production on factory-floors and tele-medicine/eHealth initiatives.
Alan explained that Proxim believes that HDTV is going to be the technology of choice for video compression because users will be demanding higher quality video images. Furthermore, Proxim thinks that that the wireless communication networks which convey the video streams are best built in a point-to-point and point-to-multipoint topology, rather than (WiFi) mesh which has fallen out of favor. He noted that Proxim’s broadband wireless transport systems that operate over these point-to-point and point-to-multipoint topologies do so over a private network (as opposed to connecting via the Internet like Cradle’s systems do, covered later in this article). Moreover, 95% of Proxim’s installations use fixed broadband wireless (both fixed WiMAX -IEEE 802.16d-2004 and a proprietary technology to increase speed and/or distance) rather than mobile broadband wireless connections.
Alan’s talk elicited two questions from the audience – the first questioner inquired why analog video surveillance technologies have found favor in practice with deployment and why digital video surveillance technologies were placed on the back burner after seeing initial deployment. In his answer, Alan pointed out that digital video surveillance technologies need high-quality digital cameras, but also require a reliable transmission network (wired or wireless) which can provide steady bandwidth to transmit the video surveillance data to a point of aggregation like a central video server. In the absence of sufficient constant bit rate bandwidth, the resulting digital video stream quality will be unacceptable due to jitter or freezing of the image (caused by an empty playback buffer). The lack of sufficient network bandwidth was a major cause of digital video surveillance technologies failing to gain a large market share compared to analog systems. The second question related to the impact of electromagnetic interference (EMI) on the video information. Alan explained that the new wireless broadband communication systems (both WiMAX and LTE) employ a multicarrier modulation scheme such as orthogonal frequency division multiplexing (OFDM) which is fairly resistant to EMI. Furthermore, OFDM can also be combined with multiple input multiple output (MIMO) transmission schemes to minimize the likelihood of errors at the receiver end.
In his talk on “Video Surveillance, Security and Smart Energy Management Systems,” Suhas Patil explained that the recent improvements in semiconductor chip set capabilities and new computer architectures have promoted the growth of digital video surveillance technologies such as those which employ network video recorders to aggregate video from an entire city. He also pointed out that a key contributor to the adoption of video surveillance systems in many parts of the world (despite concerns of invasion of personal privacy) was the possibility of, and actual occurrence of terrorism with the city of London (U.K.) being a pioneer in this regard. Suhas, while describing the structure of the video surveillance system, noted that a critical requirement for these systems is that they be resilient to any fault at anytime. These faults can include network breakdowns, power failures, control room disablement or faults caused by extreme ambient temperatures or extreme climatic conditions. As far as access technologies are concerned, digital video surveillance systems can use WiFi mesh networks, WiMAX networks or proprietary communication systems. According to Suhas, the technology underlying a digital video surveillance system is a highly complex one, employing state-of-the-art hardware such as cameras, storage systems and servers and state-of-the-art software including operating systems, and transmission technologies. Thus, the entire system needs very careful design in order to maximize it’s efficacy. Suhas also briefly explained videoanalytics as a system which can detect an object such as a human being behaving in a manner which would be difficult for a human observer to notice, such as a party guest moving around the room in a random or rapid manner compared to other party guests. Finally, the talk then pointed out several major challenges faced by video surveillance systems including the need to keep abreast of the rapidly changing technologies as well as practical deployment challenges.
In a response to a query on working on still images in a pixel-by-pixel fashion, Suhas explained that it is possible for a basic camera to capture raw information about a scene and then have the data processed pixel by pixel in order to maximize the dynamic range before the image is converted to a JPEG format. During the Q&A, Suhas briefly recounted the history of his company, Cradle Technologies, as being a spinoff from Cirrus Logic and which built multicore processors long before anyone else thought the technology (multicore processors) as being valuable. Additonally, while clarifying his assertion about digital versus analog video surveillance technologies, Suhas noted that while analog technologies are better suited to low-light conditions and offer better dynamic range than their digital counterparts, digital signals allow for better image resolution than analog signals. However, it is difficult to claim one technology as being clearly superior to the other.
The final presentation of the evening, by Basant Khaitan titled, “The Role of Video Analytics in Video Servers and Network Edge Products,” explained the nature of video analytics (VA) as a young field which has also been referred to as video content processing or video content analysis. Basant explained that VA can be defined as the real-time classification and tracking of objects like people and vehicles by using their (objects’) outlines rather than any bodily or facial features. The analytics system can be either co-located with the camera itself (at the network’s edge), or situated at a central server which receives the video streams from the various cameras at the network edges. Additionally, Basant pointed out that while a video frame’s size is of the order of megabytes, the corresponding analytics information for that frame is often no more than a few hundred bytes in size. While explaining the technical details of VA, Basant opined that modern VA systems produce results which are sufficiently reliable for practical use despite the presence of artifacts born from poor ambient light or dust-filled air. Basant then elaborated on a practical VA system built by his company and currently being used by the police in Calcutta, India for traffic management. According to Basant, VA systems built for such purposes as traffic control are highly mission-critical and require 100% reliable operation. In such conditions such as those found in developing countries, VA systems face severe challenges due to the presence of dense populations and poor public compliance with traffic laws. Furthermore, in tropical countries, extreme climatic conditions such as hot weather, rain-flooded streets and dust-filled air can also hamper the quality of the analytics results. In the case of Calcutta, all of the above conditions are faced by the VA system which is being used to control the city’s traffic lights. In this case, network-edge deployed analytics information is sent to a local video server (which, incidentally, was developed in cooperation with Cradle Technologies) from where the information can be remotely retrieved and viewed. Basant pointed out that all several intersections of Calcutta are now monitored by the VA system which has now replaced the previous system which was monitored entirely by human beings.
Panel Discussion Q and A:
The panel session was more of a collaborative Q and A with Suhas Patil and Basant Khaitan. There were many questions from the meeting attendees, several of which could not be entertained due to a shortage of time (see list of unaddressed questions below).
In a response to a question on the number of video surveillance cameras which are connected wirelessly vs wireline connectivity, Basant mentioned that none of the cameras in their deployed systems are connected wirelessly at this time, and that all camera connections are of the wireline type. Suhas noted that while most cameras are connected via a wired connection like a CAT5 cable, the access to the content can be accomplished wirelessly via a cellular service (such as in the Indian scenario) or by a WiFi connection. WiFi-based access to the video server is also available if the server is connected to the Internet (Cradle’s server is one such example). The sensors in the video surveillance network themselves can also be connected wirelessly via a ZigBee mesh network, although they have not yet been so connected.
A question was raised about whether video processing is ready to see any major innovation such as what Map Reduce technology did for Google’s text processing. Suhas responded by explaining that the video analytics for applications like license-plate recognition can be done on the cloud. When queried about how Cradle’s technology can help mitigate the impact of, or altogether prevent future terrorist attacks, Suhas pointed out that the contemporary video surveillance systems had either failed altogether or had malfunctioned during terrorist attacks. Cradle’s system, on the other hand, continually monitors the deployed system for functional integrity via a central server cloud in order to ensure that it (the video surveillance system) is fully operational at all times.
The panelists were then asked a question about what is the preferred mode of connection for a city-wide array of cameras. Suhas invoked the example of Cradle’s approach to digital video surveillance, where fixed broadband wireless access via WiMAX or WiFi mesh is used to connect their networked video server to the Internet. Furthermore, a IP-VPN client (such as a PC or other screen-based device) is connected to the networked video server through the public Internet via a 3G or mobile WiMAX connection. The panelists, in a response to a question regarding the need for video analytics in countries such as those in Asia where labor is abundant and cheap pointed out that since an average human being’s concentrated attention span is only about 10-15 minutes, and the fact that, for the overwhelming majority of the time, nothing significant occurs which warrants raising an alarm, it is imperative that an automated VA system be put in place for such applications as were mentioned earlier in this article.
The final question of the evening inquired on how real-time bandwidth fluctuations within networks such as the mesh networks affect the VA performance. Suhas Patil mentioned that by placing the video server as close to the network drop-off point (i.e., close to the camera) as possible allows for good quality video to be streamed to the server. Thereafter, special network access techniques which circumvent the fluctuating bandwidth can be used to remotely retrieve the video information stored on the server.
After the panel session concluded at 9pm, several attendees stayed on for one on one interaction with the speakers. This continued till 9:15pm when the lights were turned off and we were forced to vacate the auditorium
Unaddressed questions (submitted by Alan Weissberger to ComSocSCV Discussion List):
- Where is video surveillance used now and what are the driving applications?
- Are most of the video surveillance network architectures fixed point to point or point to multi-point, rather than mobile/wireless broadband?
- What role will 3G (EVDO, HSPA), WiMAX (fixed and mobile), and LTE play in delivering video content? Why is mobile broadband required for video client access?
- Are proprietary wireless technologies more cost effective for the performance they offer? Is this a concern for the customer?
- What type of security and network management is being used in video surveillance systems, e.g. for authentication and to prevent intrusion or monitoring?
- What role does video analytics play to augment the potential and power of a video surveillance system? Can it also be used as a stand alone offering?
- Why are IP VPNs needed to convey and deliver the video content? Why not use a dedicated private network instead?
- Is there any intersection between high end video conferencing and video surveillance systems? Are the same cameras, video transport facilities, and network management used for each? What are the key differences?
- What new technologies or business models are necessary for video surveillance to become a really big market?
- What are the current barriers/obstacles to success are the video surveillance and video anaytics markets now experiencing?
- How have terrorist attacks (e.g. Mumbai attack in late 2008) and national disasters (e.g. earthquakes) effected the video surveillance market? What is the opportunity here?
Comment and Analysis (from Alan J Weissberger):
1. Proxim’s answer to the question, Why Video Surveillance? Included these bullet points:
· Perimeter, public monitoring solutions are becoming a key component for enterprises
· Educational, healthcare and financial institutions are beginning to rely on surveillance systems to ensure safety within their premises
· Public safety organizations depend on archived data from video monitoring systems to reduce vandalism in troubled neighborhoods
· Live traffic surveillance is increasingly being used as a tool in community protection
· Terrorist threats and public safety challenges continue to drive the need for high quality remote surveillance and timely response
Additionally, we’d include production plant and factory floor (remote) monitoring to prevent schedule slips and ensure good quality control.
2. The role of broadband wireless networks in stimulating video surveillance:
-Fixed broadband wireless point to point and point to multipont networks and equipment (e.g. Motorola Canopy and Proxim’s products) that replace equivalent topology wireline networks for delivering video over a private network. Both proprietary fixed broadband wireless technology or IEEE 802.16d fixed WiMAX are used.. Those broadband wireless networks cost a fraction of the equivalent wireline networks and can be provisioned in a much shorter timeframe. Fixed WiMAX could also be used to access the broadband Internet in an IP VPN scenario.
-Mobile broadband (3G+, mobile WiMAX, LTE) which adds a whole new dimension to video surveillance and enables many new applications, e.g. IP VPN mobile client observing video images in remote locations, cameras in police cars transmitting video to police HDQ building while moving at high speed, emergency vehicles transmitting videos of natural disasters (hurricances, earhquakes, etc) to 1st responder locations that will deal with the problem(s).
3. It’s important to distinguish between the broadband wireless network architectures and topologies of Proxim (a wireless broadband transmission/ backhaul company) and Cradle (a Networked video server/client solutions company) :
a] Proxim makes broadband wireles transport systems that operate over a pt to pt or pt to multi point PRIVATE network. Those systems backhaul video surveillance and other traffic to one or multiple destinations. Proxim says that 95% of their installations use fixed (rather than mobile) broadband wireless connections.
b] Cradle uses fixed BWA (Wimax or mesh WiFi ) from their Networked Video Server to access the Internet. On the client side, Cradle uses 3G or mobile WiMAX to connect the IP VPN client PC or other screen based device to the Networked Video server through the public Internet. The key issues with that approach is that the end to end IP VPN server to client connection has to be high bandwidth and near constant bit rate, while the client access needs a high bandwidth, steady state mobile broadband connection to observe the MPEG 4 coded video over the IP VPN connection while in motion. Otherwise the video image will be unacceptable or freeze.
4. Basant’s example of controlling Calcutta traffic lights using video analytics integrated with a Networked Video server was a great demonstration of the underlying technology and proof of how valuable it is.
Here are a few background articles on video surveillance and analytics:
Video Surveillance and WiMAX- a great marriage or not meant for each other? Four companies weigh in! (all 3 speaker/panelists+ Sprint were interviewed for this article)
The Wireless Video Surveillance Opportunity: Why WiMAX is not just for Broadband Wireless Access by Robb Henshaw
Video Surveillance Going Fwd, Suhas Patil, ScD
Remote Access Video Surveillance & Analytics, Cradle Technologies
INTELLIGENT VIDEO ANALYTICS, a Whitepaper
Exclusive Interview: Robb Henshaw of Proxim Wireless!http://www.goingwimax.com/exclusive-robb-henshaw-on-proxim-wireless-5857/
Video Surveillance Product Guide