Bloomberg Businessweek: To Catch a Thief - on Video

December 12, 2005 | Coverage

Bloomberg Businessweek

California startup 3VR is among those devising new ways to quickly and smartly scan hours and hours of security-camera records

By Arik Hesseldahl

Running a hotel in a big city carries a unique set of security challenges. Chief among them: Troublemakers have no shortage of places where they can, well, make trouble. Troy Strand, manager of Chicago's Talbott Hotel, was reminded of this recently when a guest's room was burglarized and the intruder took a purse with money and sunglasses.

Police arrived quickly and began reviewing security-camera footage. It revealed images of a woman in the elevator rifling through a purse, removing items, then tossing the empty bag out the elevator door and onto another floor. Turns out she was a prostitute who had earlier accompanied a hotel guest to a different room.

"POLICE WERE SHOCKED."  Yet the most astonishing facet of the footage wasn't what it contained, but how it was searched. Typically, it can take hours of painstaking review to catch someone in the act of a crime. But in this instance, a hotel staff member used a desktop computer to call up a collection of clips of the perpetrator as easily as one carries out a Google search. "We were able to show them several angles taken from several cameras with the loot," Strand says. "The police were shocked when they saw how much we could show them in so short a time."

Security-camera systems are in some ways similar to a typical TiVo (TIVO) set-top box. Though much bigger, they store streams of recorded video on hard drives, constantly recording activity at doorways or hallways, or sensitive areas such as supply rooms. Yet while a TiVo is smart enough to let you search stored recordings for episodes you want to watch at any given time, security-camera systems can't easily reveal, say, how many times a particular person entered a given room in a specific time frame. Or they can't readily detect when an item was removed from a room immediately after a person entered and left that same room.

But that's changing. With help from technologies as varied as facial recognition and motion analysis, the thousands of hours of security-camera footage stored on hard drives around the world are becoming increasingly searchable. That's opening a whole new set of possibilities in the arenas of building security and law enforcement.

HUGE CHORE.  Only three weeks before the burglary, the Talbott had installed a new system from a San Francisco startup called 3VR Security. The setup includes a rack-mounted box that was connected to the hotel's existing system of digital security cameras. Once connected, it made the footage shot by the hotel's network of 48 security cameras searchable.

3VR CEO Stephen Russell says the box runs software that analyzes video footage and catalogs it based on what happens, who's in the frame, the time of day, location, and scores of other conditions. "Every organization that uses video-surveillance technology is likely overwhelmed with the job of staring at screens looking at live feeds all the time and then searching through hours and hours of stored video," Russell says.

When it comes to sifting through scads of video from myriad cameras, time is of the essence. In the wake of the London subway bombings last July, investigators had to pore over hours of recordings from some 2,500 individual surveillance cameras.

WELL-TIMED TECHNOLOGY.  The global market for video-surveillance equipment was $5 billion last year, and cameras accounted for more than half of that, or $2.7 billion, says Frost & Sullivan analyst Karthik Nagarajan. That market's growth will only accelerate as more cameras become connected to computer networks and are increasingly controlled remotely over the Internet. Frost & Sullivan estimates that the market for software that complements these systems is worth $153 million, and it projects growth to $670 million by 2011.

3VR doesn't disclose sales figures and operates in a niche of the market. Still, Nagarajan says, "3VR has some promising technology, especially as more companies get into video analytics and video management," he says. "This is the right company at the right time."

Russell is a former vice-president at Inktomi, the Web-caching outfit that's now a unit of Yahoo! (YHOO). He heads a team that includes Bob Vallone, a former engineering vice-president at TiVo, and Tim Ross, a marketing executive who has worked with or for Hewlett-Packard (HPQ), Adobe (ADBE), Seagate Technology (STX), and Time Warner's (TWX) AOL. Besides the Talbott, a small boutique hotel in Chicago's Gold Coast area, other customers include pharmaceutical giant GlaxoSmithKline (GSK) and Johnson & Johnson (JNJ). 3VR recently landed an investment from In-Q-Tel, the investment arm of the U.S. Central Intelligence Agency.

PROTECTION OR INTRUSION?  Russell describes the information the 3VR system gathers as a "catalog of human activity." When nothing is happening in front of a particular camera, it records nothing. But when someone appears, cameras snap to life. As the footage is fed into the hard drive, it's analyzed. The system builds an index of related information, logging such data as what time you came through the front entrance, which elevator you took to your office, or what time you were last seen in the break room or near the supply area.

Even after the footage itself is deleted, an information log created from the video remains. Russell says the next iteration will get even smarter. "In our second version, we focused a lot on recognizing people and their faces," he says. "For our next version, we're working to detect where you walked and whether or not you took anything from a given environment."

And while privacy advocates may complain about yet another way to keep tabs on people's comings and goings, Russell notes that 3VR's technology allows security personnel to search only for the footage they need. It aids in the monitoring of people doing things they shouldn't be doing, focusing less on images of innocent people doing nothing suspicious. "Today, if you're trying to find an image of a thief, the only way to do it is to pull up all the raw video and watch the whole thing," he says.

HAPPY OUTCOME.  Moreover, anyone who doesn't like the idea of surveillance cameras watching a growing number of moves and activities had better get used to it. One study by Frost & Sullivan pegs the number of cameras in use in London alone at 4 million, or 1 for every 14 people.

Russell says the system's cost runs about $2,000 per camera. Talbott manager Strand says the hotel has plans under way to double its current number of cameras, saying, "We're adding 48 more."

The results in this case speak for themselves. After calling up the footage of the purse-snatching prostitute, police were able to arrest her within hours and recover the property. "The guest got everything back," Strand says. "That's as good an outcome as you can expect." Would-be wrongdoers are advised to take note. 

Read Article on Bloomberg Businessweek