Video Analytics 2009 Mid-Year Market ReviewBy John Honovich, Published on Jul 12, 2009
The last six months were rough for video analytics. Little good news and lots of public recognition of video analytic troubles. In our own recent analysis and survey on video surveillance technology trends, video analytics finished a distant third to megapixel cameras and hybrid NVRs, barely beating "Other."
There were very few new product announcements, especially compared to the landslide of new products in megapixel and hybrid NVRs. The biggest two 'new' video analytic products are probably BRS Labs and Cernium's Archerfish. As we looked recently, lots of questions surround BRS Labs claims. Archerfish poses less technology questions but as a $1,000+ USD residential/SMB product, it will be interesting to see how well it can do for consumers spending less and staying home more.
On the other hand, the products certainly are not getting worse. Indeed, there is good reason to believe that the products are slowly getting better with more time to optimize software and faster DSPs/CPUs available.
Nonetheless, the sins of video analytics certainly have caught up to them. The trade magazines and the analysts have now both acknowledged and recognized the dissapointments of video analytics.
Examining Video Analytics for Fire Detection
A webcast on Fire Detection Video Analytics [link no longer available] from ISC West has also been released. This one is from AxonX, one of the larger specialists in this niche. Fire detection is one of the less frequently discussed analytics.
The prime use of video analytics for fire detection is within the commercial/industrial markets. For residential and regular offices, traditional fire detectors can cover the same area as video analytics for far less cost. By contrast, in warehouses, stadiums and other large areas, traditional fire detectors are cost-prohibitive and infrequently used.
Video analytics can cover large areas from far away. AxonX claims to cover 20,000 square feet. Unlike traditional fire detectors that require smoke or particles to reach the detector, video analytics can detect smoke and fire earlier in the process.
Key to the deployment of video analytics for fire detection is appropriate insurance or standards approvals. In the US, AxonX cites the FM-3260 as a key standard for compliance.
While video analytics for fire detection has potential, it is clearly limited to a narrow section of the market.
Waterside Video Analytics Whitepaper from Acuity
Acuity has issued a new whitepaper on the use of video analytics for waterside usage [link no longer available]. Deploying video analytics near water is one of the more difficult uses as the motion and reflectivity of water makes analysis more challenging. The white paper discusses a number of the environmental and logistical issues. Also, importantly, it cites false positive rates for each scenario, a necessary element for any statement on video analytics' performance to be useful.
IMS Statistics on Video Analytics
More acknowledgement of video analytics poor performance this week, delivered by IMS:
- Only "moderate growth" projected this year and next
- 30% annual growth rate by 2012-2013
- 2008 sales volume: 38,000 channels server based, 40,000 channels edge based (cameras, encoders)
- Government and Transportation about 50% of total video analytics market
- IMS states, "The market is still awaiting that “killer application,” and no one seems certain yet what that will be"
Vidient Company Update
Over the last year, Vidient reports that it has been focusing on reducing false alerts and enabling simplified deployments (new version 4.0 will be publicly announced soon). As evidence of their improvements, they cite a 300 camera deployment for the St. Louis (US) metro transit service. At this deployment, they claim 97% accuracy with only 1 false alert per camera every 3 days. They mention an article in GSN magazine. I have asked GSN for a link to an on-line version and will share when I get it.
Vidient reports a new financing round (in the last few months). They project that this round is good until at least 2010 when they expect to be cash flow positive.
Vidient's analytics are still server based (no smart cameras, TI Da Vinci port etc). Their rationale for this is that smart cameras do not provide sufficient computing resources to implement filters and enhancements to meet performance needs.
Finally, Vidient reports adding a feature that allows users to flag false alerts in production systems, enabling the analytics to adapt to specific on-site conditions. Vidient says that this is in addition to algorithmic improvements.
Agent VI analytics Certified by i-LIDs
Agent VI announced that its analytics have been certified by i-LIDs [link no longer available]. i-LIDS [link no longer available] provide the UK government's benchmark for video analytics. i-LIDS offers a variety of certifications.
Agent VI reports that is certification is for primary detection in sterile zones. From what I know of i-LIDS, there is primary and secondary detection. Primary is more impressive because it indicates that the analytic can be used by itself. Note: that it is certified for use in strerile zones which can be an issue for many deployments where movement will occur.
There's a debate about the meaning and important of i-LIDS certification. My understanding is that it is important for UK projects (as it is a necessary requirement for some projects).
Video Analytics for Remote Monitoring
Remote video continues to be a popular topic for video analytic vendors. In the past few weeks, we have seen MACE advocate remote Video in a trade magazine article [link no longer available], ioimage announce a project where they reduced guard costs 40% for an Israel car dealer [link no longer available].Moreover, VideoIQ recently conducted a webinar on remote video where they discussed a 1000 camera deployment (see VideoIQ's slides [link no longer available]).
While VideoIQ claims that a single operator can monitor over 1000 cameras, the integrator on the webinar noted that their deployment in Birmingham Alabama had significantly below that ratio. Separately, VideoIQ reports an average of 1 false per 3 days per camera.
These deployments still seem early stage and do not reflect wide spread adoption. There are a number of barriers (beyond false alerts), like integration with central monitoring station software and the costs of adding video analytics (almost no existing system supports video analytics without adding new hardware on site).
Nonetheless, remote monitoring is the most promising area for video analytics as it offers hard cost reduction (reducing guard costs) that security organizations find easy to justify.
Limitations in Facial Recognition and LPR
While facial recognition for identification verification (like Driver's Licenses) continues to expand, the technology suffered a PR black eye this week. A number of US states are now banning smiling for driver's license photos. Unsurprisingly, the public and the mainstream press was not happy about this. Summing up the sentiment, a popular gossip website deadpanned, "Facial recognition software operates on the principle that people never make expressions ever, and just wander the earth looking like dead-eyed zombies." Obviously, that's not technically true but it does express a significant limitation in today's analytics - differences that appear minor to humans can be very hard for computers to discern.
This is not unique to facial recognition as a similar issue exists with License Plate Recognition. In selecting LPR, you need to verify that the LPR system has been optimized for the license plates used in your area (size, shape, coloring, separataion of letters, etc. varies across the world's license plates).
While there are certainly techniques in development to automatically compensate for these issues, they are not something found widely in production deployments (especially at reasonable prices).
Overview of VCA Technology
I spoke with VCA Technology this week -- a relatively new entrant in the video analytic market (UK based company founded in 2007 with core team members previously at PI Vision). VCA has the same fundamental business model as ObjectVideo yet offers a lower cost alternative. Like ObjectVideo, VCA OEMs their analyticssupporting TI DaVinci and Intel chips [link no longer available]. They offer two levels of video analytic offerings, with the higher level offering a variety of analytic 'flavors' common in the market today.
While I received a demo of the product, I have not tested the product. I found their 3D configuration tool to be quite useful. However, certainly I cannot speculate on the performance of their analytics.
Using Video Analytics to Cut Guards
One of the best security management blogs ran a series of posts this week on the use ofvideo analytics to eliminate guards. The application is secure facilities where US TWIC regulations require that visitors without clearance be escorted.
As I describe in a comment to that post, this is a risky way to cut costs. There are problems with saturating coverage with cameras, maintaining 100% accuracy on tracking and delays in responding to threats. Making this worse, the subjects are already inside the premises, making it far easier and faster to cause damages.
In this case, it's Siemens using the Vistascape product at the Port of Wilmington. Siemens praises the ability to use the Port's existing CCTV investment. However, you have to wonder what the security and operational risks really are.
However, the most common use of video analytics to reduce guards are usually in construction sites or moderate security areas, which pose far less security risk and are significantly simpler than deploy than the TWIC escorting scenario.
Overview of Acuity - New IP Video Manufacturer
Acuity [link no longer available] is a new video surveillance company (1-2 years old) founded by Pelco and AD veterans gaining recent press attention. I received a company overview and live demo from them this week. Here are my initial observations:
Acutiy provides video management software and NVR appliances running only on Linux. They are positioned themselves to compete against IP video surveillance providers such as Genetec, OnSSI and Milestone. I see 3 primary ways they are trying to differentiate: (1) providing their own suite of video analytics, (2) deeply integrating the use and management of video analytics in their user interface and (3) providing security oriented feature and performance enhancements (that leverage their Pelco and AD experience).
Since this was an introductory call, I do not know the specifics nor have I tested the product. The key things to examine in considering Acuity are (1) given that the product is fairly new, how mature and broad are the feature offerings relative to significantly older products and (2) as with all analytics how well do they perform in the field.
Remote Monitoring with Video Analytics
ioimage recently released a webinar detailing their entrance into remote monitoring services [link no longer available]. The webinar is a general introduction to video analytics and high-level benefits of remote monitoring.
The webinar noted some things video analytics cannot do, such as detecting erratic behavior, nervousness,pickpockets,deer from people, distinguish friend and foe.
In the webinar, ioimage offers a special discount of $999 for their IP smart cameras and $250 for encoders. This includes training and is positioned to motivate integrators to expand their use of video analytics.
Video analytic manufacturer movement into remote monitoring continues to increase with VideoIQ, Cernium and now ioimage all driving offerings.
Another Important Sign that Video Analytics is Failing
When trade magazines run negative articles, you know a technology is in trouble. This month, SDM released a report on security integrator's experiences with video analytics. It is an excellent article that actually investigates the issue rather than simply prints quotes from the marketing people of each vendor.
In the article, the most interesting claims are (1) video analytics can cost up to $100,000 USD per camera, (2) expect one false alarm per camera per hour and (3) that on a scale of 1 to 10, the difficulty of video analytics range from a 6 to an 11. A senior engineer from Adesta states that at first the difficulty of using video analytics was a 10 but after a few years of experience, it's now 'only' a 6 or 7.
How motivated will people be to buy video analytics whern they hear integrators say that it is extremely expensive, highly error prone and really hard to deploy?
Now, even I think these claims fall more on the worse case scenario and not realistic of the best that is available. However, it is very clear to see the level of problems that video analytics faces. I see no technical solution in the short term. What's scary to me is that most video analytic vendors are not even making claims anymore about how they are resolving this. Even the video analytic vendors tend to join the chorus of establishing 'lower expectations.'
Cernium Expanding into QSRs
A new job posting by Cernium indicates that they will be selling their new Archerfishvideo analytics appliance to Quick Serve Restaurants. They are looking to close franchise agreements with large fast food companies like McDonald's and Subway.
QSRs are well known for being cost conscious on video surveillance so it will be interesting to see how Cernium builds their value proposition. The early marketing around Archerfish was about homeowners/small business people getting alerts when they are away. See an interview with Cernium's CEO [link no longer available] where he indicates current pricing is $1495 for a 4 channel analytic appliance with 2 cameras.
Will this appliance become a DVR? I think it is going to have to if they are to sell in the competitive QSR market.
Does Bosch Over-Sell Analytics?
Bosch has a new video analytics 'game' this year. Stealing a drill has been replaced with going across a 'football field.' I questioned whether Bosch was over-selling video analytics. The ensuing discussion is an interesting look about what analytics are supposed to do and what constitues over-selling.
Warning on Avocado Security: Avocado is a startup claiming to do a combination of video analytics, Software as a Service and Business Intelligence. They have been around for a few years but recently started a new marketing push. A few months ago, I researched and did an interview with Avocado. I think the company is significantly misrepresenting their abilities. Please be careful about partnering, selling or using this product. It's conceptually appealing because they claim to have put together all the buzzwords in the industry but the company was completely incapable of describing how they were able to make it work. A key example is their claimed 'universal compatibility with 3rd party systems' (described in a Frost 'movers and shakers' piece). They claim to work with any DVR including citing by name many of the industry leaders. I checked with multiple DVR companies and none verified any integrations. Anyone who knows how VMS technology works knows that universal compatibility is impossible in today's industry. There were other red flags. E-mail me if you want to discuss details.
ioimage looking to acquire companies
This week, the founder of ioimage, Roni Kass posted a note on the LinkedIn Video Analytics Group [link no longer available]. His note stated, "We are interested to collaborate/acquire different companies, which have complementary technology/products to our products. These can be unique intelligent video algorithms, special sensors, and unique hardware solution. Companies may be pre-revenue (in advanced prototype stage), initial revenue, or more mature companies." In an e-mail exchange with ioimage, Kass stated, "Business is going well, good time to grow, and we cannot do everything in-house". He did not disclose funding available for acquisitions. I think it's most likely that any acquisition is more for technology than established brand as most brands in video analytics are too weak to have value and the few that do would cost too much. Also interesting is the use of a social network like LinkedIn to prospect. It's a sign of the increasing power of the Internet to conduct business. And because LinkedIn provides a global scope (that group has almost 2000 members), ioimage may be able to find a relatively unknown company without spending on traditional middlemen.
Vistascape/Siemens SiteIQ Update
Siemens contacted me this week to give me an update on the Vistascape product, now named Siemens SiteIQ. It looks to be largely the same product from the Vistascape era. They are claiming improved accuracy and benefits of longer range detection (similar to SightLogix). The product is being positioned to larger-scale deployments (20 cameras or up). The product still requires servers, prefers analog cameras and does not run on smart cameras.
Mango / Mate merger
Israel article reports encoder manufacturer Mango DSP and analytics provider Mate are merging [later confirmed]. Both have suffered fairly significant layoffs recently. The Israel report says that the companies have joint investors and that this is a way to reduce costs. Beyond that, I am not sure about the product 'synergies' as you can run MATE on Mango anyway. Video analytics is in for tough times.
Future Video Analytics
This week was the IMS conference and Geoff from SecurityInfoWatch reported on video analytics progress. Money quote: "an industry niche that is maturing, but is still somewhat in its infancy." What's strange is that this is the same thing people said 5 years ago. The argument is made that the industry is not 'in dire straits' citing many deployments but then cites that only "$50 million was spent on PC-based (server-based) VCA projects in 2008." For a market segment with hundreds of millions of investment starting from the first half of this decade, it's a strange situation. Kohls' report talks about the problems of overselling but does not talk much about technological advances. However, a recent article fromSecuritySystemsNewsEurope does, citing IBM and Milestone. IBM talks about "cross-correlation of analytics happening on different, multiple camera views" which sounds great except it's not feasible to do in practice at any reasonable cost. Which, of course, brings up back to more over-selling.
Axis position on video analytics
Axis is advocating using analytics on IP cameras. This is not news but I was surprised to hear their questionable rationale (see the 3 minute video). Axis claims the main problem for video analytics is that the price is too high for running analytics on central servers. To the contrary, the industry consensus is that analytics' main problem is poor performance - whether on servers or cameras. Indeed, performance historically was better on servers simply because of far greater processing power at far lower cost. Poor performance is a key risk of using analytics on existing IP cameras. The most accurate analytics in the market all use either purpose built cameras (like ioimage or sightlogix) or they use specialized chips optimized and dedicated to video analytics (like ObjectVideo on TI chips). The problem with Axis' approach is that running analytics on regular IP cameras greatly increases the risk of performance problems. Axis often cites via:sys and AgentVi as two examples of analytics that run on their cameras. However, my tests of via:sys show poor performance. AgentVi tends to work better but, of course, still requires a central server. Also, when loading analytics onto existing IP cameras, you need to be very careful about processor overload. You can easily cause performance problems by simultaneously using advanced compressions, high performance, multi-stream and video analytics at the same time. In this interview, Axis claims that with IP cameras you can "take any analytic from any analytic vendor." This is only true conceptually. The practical issues I outline make this very unlikely with today's Axis cameras. Quite frankly, I am dissapointed with Axis' stance. One of the things that I respect Axis about highly is their commitment to rolling out products that work really well and about the care they take in not endorsing science projects. Unfortunately, this interview makes assumptions that are technically quite questionable.
Video Analytics for Remote Site Monitoring
A test of a manufacturer, Emza, who just focuses on intrusion detection video analyticsfor remote sites. It worked well and I like the focus. Narrowly targeted products tend to do better as the company can really optimize the functionalities and cost around that niche.
AgentVi on Sony
AgentVi announced that Sony's IP cameras will now support AgentVi analytics. AgentVi's approach to analytics is different from well known providers like ObjectVideo and ioimage. AgentVi splits the processing doing some of the work on the camera and the rest on a server. Because of this, AgentVi can work on a variety of common IP cameras including many Axis cameras and now Sony. By contrast, ObjectVideo's analytics are constrained to cameras with higher end chips, thereby reducing the cameras that can run their analytics (e.g., OV is not supported on Axis nor Sony). Of course, ioimage only runs on their own cameras and encoders.
Intellivid / AD: Last year, Tyco bought Intellivid, a specialist in video analytics for retailers. This week, American Dynamics, Tyco's video surveillance company released their Intellivid datasheet. It has a lot of interesting information. It seems that the analytics will run on a central server and will analyze and record video from IP cameras.
MATE has announced a new version of their analytics as well as integration with Hirsch. Given the severe downsizing MATE recently went through, it will be interesting to see if MATE can overcome the current economy.
Vidient Red Herring Top 100
Recently, video analytics vendor Vidient announced that it received a Red Herring award. Vidient has been struggling for a long time (and to be fair so have most of their peers). But a Red Herring Award is joke. Not only does Red Herring have little idea about the value of such products, Red Herring seems to have problems staying in business itself. These awards mainly serve to mislead and confuse buyers.
Cernium announced their new residential monitoring product - Archerfish. I find this risky - both because of the risk in making this work for residential customers but also the impact it has making Cernium even less focused.