The Camera is Watching You: VideoIQ Puts Smarts into Surveillance
Even before events like 9/11, the Madrid train bombings, and the London Underground bombings, governments and corporations were busy blanketing outdoor and indoor spaces with networked cameras, reasoning that video surveillance can help security professionals spot criminal activity in progress, or at least get it on tape. But there’s one big problem with the blanket approach: the more cameras an organization installs, the more people it takes to watch them, and the more recorded video there is to review if an incident occurs. It’s a classic case of information overload.
Many analysts have predicted that this problem will eventually be solved by moving some of the intelligence in a video surveillance network to its edges: that is, by building “smart” cameras that can determine on their own whether the events they’re observing are suspicious, then alert human operators in real time, or store the video stream differently, so that there’s more information to help investigators later.
And now a Bedford, MA, company called VideoIQ—a 2007 spinout of General Electric’s security division—is taking the first steps in that direction. The company is set to release its first smart digital surveillance cameras in August, and it announced today that it has more than doubled its venture financing pot, raising $10 million in a Series B round led by Lehman Brothers Venture Partners. (Existing investors Matrix Partners and Atlas Venture, both headquartered in Waltham, MA, were also in on the round.)
VideoIQ’s new camera, the Intelligent IP Surveillance Camera with Video Recording, or iCVR, has a built-in Linux computer with video analytics software that can identify events of interest—for example, a person approaching the perimeter of a power plant—and notify a security guard. It also has a hard drive that’s smart enough to save digital video data at full resolution when there’s something interesting going on within the camera’s field of view, but compress it when there’s nothing happening, meaning that it can store up to two months’ worth of data, compared to the week or two stored by many older systems.
“This is the first generation of a very powerful technology that we believe will revolutionize surveillance, because the vast majority of all video systems are not monitored at all,” says VideoIQ president and CEO Scott Schnell, a former Atlas entrepreneur-in-residence who is a veteran of RSA Security, Photonics, Apple, McKinsey, and Chevron. “The primary purpose is to enable the delivery of forensic information quickly and efficiently, if something does happen.”
The iCVR isn’t the first surveillance camera with built-in hard drive and video compression software, but it’s one of the first to add video analytics software to the mix. Currently, adding motion-detection or object-tracking capability to an existing surveillance network can cost $1,000 to $2,000 per camera, according to Schnell. But because VideoIQ’s analytics software runs on the same built-in processor used for compression and storage management, customers essentially get it for free.
And if an organization replaces or upgrades its existing surveillance systems with VideoIQ cameras—a project that will cost about $1,800 per camera, Schnell says—it also gets a distributed storage system that doesn’t rely on a central recording device (a collection of VideoIQ cameras functions, in essence, as one big networked DVR), as well as cameras with state-of-the-art sensors from Mountain View, CA-based Pixim that are capable of seeing in both bright sunlight and low-light, nighttime conditions.
The biggest initial customers for the system, Schnell believes, will be large facilities with extensive outdoor perimeters that need protecting. “The larger the perimeter, the more costly it is to have manned patrols, and the more monotonous it is” to monitor video cameras visually, Schnell points out. But if the cameras themselves are tuned to watch for for unusual events, “You don’t really have to monitor them anymore,” Schnell says. “What you’re really doing is responding to threat alerts from the cameras, and only then looking at live video.”
Video IQ has been beta-testing its system at several dozen commercial sites such as big car dealerships, Schnell says, and at some of them, “there’s one individual monitoring literally hundreds of cameras simultaneously.” But that’s no longer such a mind-numbing task, since guards don’t have to stare endlessly at a signal where nothing is going on.
The iCVR has another capability that could make it appealing to organizations that manage more active spaces, such as shopping malls. The company’s analytics software goes beyond the relatively simple task of object tracking—following a blob of pixels as it crosses a camera’s field of view—to what computer vision researchers call “object classification,” meaning it can tell a human from a bird, a tree, or a car, and even keep track of an individual person in the image, provided that he or she is wearing reasonably distinctive clothing or moving in a distinctive way.
That opens up some interesting scenarios, such as the ability to search stored video based on a person’s visual appearance. In one YouTube demonstration video created by VideoIQ, a father loses track of his young daughter in a crowded mall. Security guards are able to find a recorded frame from the moment the father and daughter entered the mall, then click on the daughter’s image and use it as the input for a search of the video data from all of the mall’s cameras, looking for objects with similar shapes, sizes, and motions.
In the video, guards quickly locate the girl in the mall’s play area. In reality, says Schnell, “We’d probably find a dozen little blond girls dressed similarly to the missing one.” But at least that would save security officials from having to review hours of videotape or scan dozens of cameras manually, greatly speeding up the search. “It’s not CSI, it’s not Minority Report,” Schnell says, referring to the way popular media exaggerate the capabilities of surveillance and video-enhancement technologies—a theme that also came up in our recent profile of Brighton, MA-based video analytics company Salient Stills. “But it is a practical technology that will find you the 20 likely needles in the haystack. And it will be up to a person to make the exact pattern match.”
Schnell says VideoIQ will use the new venture infusion (which brings the company’s total financing to $18 million) to ramp up manufacturing of the iCVR and expand the 35-employee company’s sales and marketing operation. The company will also use the funds to perfect a second product—what Schnell calls an “encoder,” meaning everything in VideoIQ’s main product except the camera. It’s designed to allow owners surveillance system owners to upgrade their existing cameras, giving them the same analytics and storage capabilities as the iCVR.
Given the size of the installed base in the surveillance world—millions of cameras covering major urban, commercial, and industrial spaces around the world—Schnell’s probably right when he calls that “a giant opportunity.” And while the prospect of smart video cameras monitoring us everywhere we go can feel both comforting and creepy, the economic logic behind automating video surveillance means that such techniques are only going to grow more powerful and widespread. “We are bringing a tremendous amount of technology to a very old industry and making that technology approachable,” says Schnell. “But we are only at the very beginning.”