It was June, 2002, nine months after 9/11, and John Romanowich was walking Ground Zero with Keith Hanna, a colleague from Sarnoff. They had been invited by the Port Authority of New York and New Jersey to discuss the possibility of using automated video surveillance to protect the World Trade Center site. “The area was completely excavated,” says Romanowich, “and was effectively a crater in the ground. All the security at the site was from roving patrols.”
Yet Romanowich saw that automated surveillance technology simply was not up to the challenge. The site was both a crime scene and an active construction site: “The infrastructure for us to work with was minimal,” he says, “even though we were in the center of one of the most powerful cities in the nation.” In particular, it just would not be feasible to install fixed cameras on the temporary fences, then run the necessary cabling throughout the site, much less install the required supporting computer equipment in the temporary trailers in use on the site.
“To my tremendous frustration,” says Romanowich, “as I walked around this facility with some of the world’s leading experts, both from Sarnoff and from other companies, we realized that this could not be done. And to this day it has not been done.”
Out of that frustration came the foundations of SightLogix, the newest business founded by serial entrepreneur Romanowich. He stubbornly kept it under wraps for five years: “Otherwise we would just have been educating the competition.”
Last November Romanowich saw his vision come to fruition when his new Princeton-based company, SightLogix Inc., held the first public demonstration of its intelligent video surveillance system — specifically designed for protecting this kind of critical outdoor infrastructure.
The demonstration was held at the Conference Center of the New Jersey Hospital Association facility, on Alexander Road in Princeton, with one live camera watching the front lawn of the campus, and a second positioned across the street to monitor traffic along a driveway and parking lot.
At the demo Larry Barfield, vice president of government programs at SightLogix, explained the attributes of the kinds of critical infrastructure locations that the company is targeting, from landmarks, to chemical and utility plants, to airports, to the nation’s borders: “They all have several things in common. They have perimeters of varying shapes and sizes, typically dispersed and remote. They also typically have a lack of infrastructure located nearby for supporting communications and power. And they are all at the mercy of the outdoor environment.”
The traditional solution for protecting an outdoor site, as at the World Trade Center, is to employ a large staff of guards to patrol around the clock. But this is expensive, and still leaves areas unwatched after the guards pass by on their rounds. And while maintaining this kind of staffing might be possible for a major central site, it just is not feasible for guarding remote locations or long borders.
Video-based surveillance would seem to offer a better solution: set up cameras to watch along each section of the site, and then feed the signals back into a control room to a wall of displays. A small staff then can watch the displays to look for unexpected activity. However, this task is not well matched to us humans. Nobody can continually watch a bank of monitors with unflagging alertness for hours and hours, much less 24/7.
Thus the attraction of “intelligent” video surveillance, to have a computer monitor the camera feeds, and generate alerts when it detects moving objects or other changes. This might seem to be a great choice for security monitoring, and in fact, says Barfield, “a number of organizations have actually tried that over the past few years.” But he says, “intelligent video, as compared to indoor locations, has had less than stellar success in the outdoor world.” (Barfield speaks from the experience of having most recently directed the security operations for all U.S. Marine Corps bases in Washington, D.C., including Presidential support facilities.)
In fact, it is very difficult to program computers to come anywhere near the natural ability of humans to detect and really understand possible threats in a scene, particularly in a busy and uncontrolled outdoor environment. Imagine a very busy outdoor scene — with grass blowing and foliage waving in the wind, flying leaves or rain or snow, dramatic lighting changes from clouds passing overhead, and even ocean waves crashing in the distance. Even with all this activity, the human eye and brain still can instinctively ignore the background noise and immediately focus in on the motion of an unwanted intruder.
However, computers only can detect the changes from one video frame to the next, and don’t “understand” the larger context of grass and trees and waves. The motion of a large branch, or the sun coming out and throwing a sharp shadow, can appear as a large moving object in the scene and set off an alarm. Or a near-by running animal can appear bigger and much more threatening than a human sneaking slowly closer in the distance.
As a result, today’s video surveillance systems are better suited to indoor environments with steady controlled lighting, and to areas with no activity, such as monitoring emergency exit doors. Otherwise, they can generate large numbers of false alarms.
“Even best in class systems can have eight false alarms per camera per day,” says Romanowich, “which is totally unacceptable. The promise was better security and reducing your operating budget, when in fact they provide marginal advances to the security while shifting the operating budget to IT services and additional staff to chase false alarms.”
In addition, installing this kind of video surveillance system for a large enterprise with tens or hundreds of cameras can impose huge logistical problems and costs. “Typically on the projects that I’ve been associated with,” says Barfield, “the infrastructure can be up to 50 percent or greater of the overall cost of the project. Intelligent video was not necessarily an attractive economic way of solving the problem.”
First, you need to install the cameras, which requires installing strong poles to keep the cameras steady in winds, plus digging trenches to run power and video cables. Then you need to run all those cables back to the guard room, and to the computers. However, this kind of computer vision processing is very processor-intensive, requiring more and more computers as you add additional cameras — typical systems can handle only four to sixteen video feeds per server. (One partial solution is to transmit the video in compressed digital format at lower resolution, which does reduce communications bandwidth and processing requirements, but also degrades the quality of the video and therefore lowers the sensitivity of the analysis — requiring more cameras and increasing nuisance alarms.)
So what might have begun as a project to deploy security cameras could turn into a large construction project, plus a major IT development effort to design and build a computer center and associated infrastructure (including cooling and backup power and 24/7 support). Again, while this approach may be possible for a major facility, it does not scale well to hundreds of cameras, much less thousands, it is not feasible for remote locations, and it does not support rapid deployment of systems to protect new or temporary sites.
So as John Romanowich walked Ground Zero, he saw that the current intelligent video surveillance systems simply were not up to the task of protecting the site. It just would not be feasible to set up the necessary infrastructure and wiring and IT systems in an active construction zone, much less be constantly reconfiguring the cameras as work progressed.
Instead, Romanowich envisioned a totally different approach, what became the SightLogix system, explicitly designed for surveillance of critical infrastructure in an uncontrolled outdoor environment.
His vision was to leverage today’s smaller and more powerful processors by building the intelligence into the camera itself. While the resulting camera unit is more expensive, the overall system cost is substantially less — providing significant savings by eliminating the cost of building a computer facility to do the video analysis, as well as the associated ongoing maintenance and support.
And this integrated approach has additional benefits: Since the processing is done directly in the camera before any video compression, it has access to the full-quality video image, resulting in more accurate results over a longer range — reducing costs since fewer cameras are required to cover a perimeter. And since the camera is self-contained, it can be totally mobile, powered by solar panels and transmitting information and associated videos wirelessly — making it possible to rapidly deploy to new locations and easy to re-configure the security at an existing site.
As a result, says Romanowich, “the capital equipment and maintenance and operation budget is significantly less.” The initial capital expense can be a third that of the competition: “One recently installed system was under $80,000 for the installed cost, and the competition was $125,000. So ultimately we covered four times the area, and at only two thirds the cost.”
While SightLogix has been demonstrating its system at trade shows and to individual customers, the November 21 event at the New Jersey Hospital Association was the first live public demonstration of the system. The event was attended by potential government and military customers including the Department of Homeland Security, Marine Corps, and Army Night Vision Labs, as well as representatives from the offices of Congressman Rush Holt and Senator Frank Lautenberg.
For the demonstration, one SightLogix camera system was positioned close to the building to monitor the NJHA entrance and lawn. A second camera was placed across the street watching the driveway of the University Park Plaza complex where the SightLogix office is located.
But the key point was that these were not permanent systems — there was no installation, no infrastructure, and no cabling. The cameras were mounted on portable trailers, positioned on retractable posts, transmitting alerts and video wirelessly to the demo display, and even self-powered with solar panels and batteries. (A third camera mounted on the roof of the SightLogix building was also transmitting to the demonstration room over a cellular network.)
The resulting feeds were projected on two displays at the front of the room. The left screen displayed the live feeds and alarm alerts from the two cameras using a third-party command and control system. Since the SightLogix system supports industry-standard MPEG-4 video format for video and standard alarm management protocols for the associated surveillance information, it can be plugged right in to security operations systems that customers already have installed. “It has to fit in to the existing security infrastructure and IT framework,” says Romanowich, “and inter-operate with all systems that might need it.”
The right-hand screen displayed the SightLogix Coordination System software for camera configuration, and for stand-alone coordinated display of information back on a geo-registered site map — telling you the two things you need to know: the location of an intruder, and its properties (size, speed, temperature, etc.).
Barfield first showed how to quickly configure the system when a camera is initially placed on a site, setting its position and orientation by clicking on an overhead view of the scene (i.e., a Google Map) so the camera can report the GPS coordinates of an intruder in real time. This is a much easier calibration process than has been traditionally required. “You are no longer required to have somebody go out into the field and have a GPS locator to locate themselves,” says Barfield. “You can do it all from the interface.”
Next, Barfield configured the alarm and display options for the camera. The SightLogix camera actually can be configured to provide up to two video feeds, with options to control the video bandwidth and overlay information. For the demo, the first feed was the raw video from the camera, and the second was processed to stabilize the image (remove shaking from wind) and to indicate detected objects by outlining them with a box and drawing a short motion trail behind them.
The camera watching the N.J.H.A. front lawn had a wide field of view, with the ability to detect pedestrians out to about 100 meters and vehicles to about 200 meters. But no one was interested in continuous alarms as traffic drove by in the distance on Alexander road, so Barfield clicked on the display to draw an alarm region that excluded the road. The system supports setting policies for different regions: active, ignore, or to use more complicated rules, such as alarming only when detecting objects of specific sizes, speeds, and/or temperature (with a thermal camera). “Boats exhibit a completely different temperature profile than the water itself,” says Barfield. “And tumbleweeds are not 98 degrees like humans.”
Policies can be more complex, such as alarming only when an object moves from one region to another. And they can be changed, for example from day to night or for a new threat level. Even better, the regions and rules are associated with the location, not the camera view, so the cameras can be repositioned without having to re-define all the monitoring information.
The demo proceeded by having a subject walk across the expanse of the lawn, moving from close to far, slow and fast and stopped, and from bright sun to deep shade (which was difficult to see with the naked eye). The SightLogix system successfully tracked the subject, as well as occasional cars that drove along the driveway on the left side, while ignoring large shadows from overhanging trees. Barfield asked the audience for suggestions, and they had the subject move off to the side and then walk back very slowly into the camera view. The system alerted very quickly — before most of the audience had noticed the subject’s leg coming into the view.
The second camera monitoring the driveway across the street took advantage of the SightLogix option to provide cameras with an IR (thermal) imager that detects heat. These work well in the daytime, but are especially useful for night viewing in locations without additional lighting. During the entire demonstration event, the camera was successfully detecting and tracking both cars and pedestrians coming in and out of the driveway, sometimes up to four cars at a time.
Barfield then had another test subject walk down the sidewalk to the left of the driveway, though a more complex background, under branches, behind light poles, and then further and further into the distance, to about 300 meters from the camera. The system tracked the subject all the way, and until they were so far away that the human eye had trouble noticing them at all.
“The SightLogix architecture allows a 3X range advantage over comparable intelligent video systems using similar optics,” says Barfield. “The performance advantage can easily translate into reduced equipment costs.” Adds Romanowich, “The competition does hundreds of feet at best with a wide-angle lens, and we do hundreds of meters.”
Not bad for a wireless camera on a stick. The camera is rugged, designed to operate in temperature environments from -30 to +60C, and nitrogen purged to ensure humidity control (with sensors to monitor these conditions). It can communicate securely (with standards-based encryption) using wireless (802.11), cellular, and SATCOM — using low bandwidth mediums without sacrificing quality, since the analysis is performed in the camera and not on the output video. And it runs on low power, so it is feasible to operate with solar power.
“Everything that was demonstrated live was performed using wireless devices,” says Barfield, “with cellular for communications and solar for power. By developing an architecture that enables wireless and solar, SightLogix has brought intelligent video surveillance to the world of mobility.”
SightLogix seems the natural evolution of Romanowich’s experience with camera designs since college. “I was seeing all these technologies evolving,” he says, “and trying to see how they come together.”
Romanowich, who grew up in Paramus, has always been fascinated by electronics and how things work. His father, a dental technician, had him playing with electronics at around age five or six. “I loved it, winding electromagnets, building buzzers, crystal radios, and ham radio.”
An active runner and swimmer, Romanowich works out daily. He and his wife, Tara, formerly a third-grade schoolteacher, have two school-aged children.
Romanowich started working with camera systems when he studied engineering at the New Jersey Institute of Technology, earning a BSEE in computer architecture in 1987 and a MSEE in optical electronics at NJIT. “I liked engineering,” he says. “In high school I was already working three jobs, at a restaurant, fixing lawnmowers, and for my electronics teacher’s small company. But at NJIT I started to get more serious.”
Romanowich earned a full fellowship from NJIT for his master’s thesis, and worked on opto-electronics and IR cameras at Sarnoff under Walter Kosonocky, an IEEE Fellow with 69 patents in solid state electronics and a key figure in the development of in-camera sensors for visible and infrared imaging. “This area is a center of excellence for camera technology and imaging,” says Romanowich, who compares it to Silicon Valley.
In 1988 Romanowich joined IBM as an engineer in Fishkill, New York, working on projects including machine vision inspection systems. He was also accepted to Columbia to start working on a Ph.D. “I wanted to start my own business,” he says. “I never saw myself as an employee.”
So he also began working on his own. “I started the business on the side,” he said, “and was making more money so I decided to quit. The Ph.D. was going to take five years and I could start a business in the same amount of time.” In 1990 Romanowich left IBM and formed his own company, Synthesis Electronics, to provide engineering consulting and services.
Romanowich began working with the Electrical Power Research Institute making advanced products for utilities. Later he returned to school part-time in the M.B.A. program at the Rutgers School of Management. “I was doing the Rutgers MBA at night with one class a semester. I got about one third done, but then I quit when I started SightLogix because I could not take money from investors and not be fully focused on the business.”
Much of Romanowich’s business was with Sarnoff and its clients. He worked on chip designs for “exotic cameras,” including infrared vision cameras for the White Sands Missile Range that were used on the Space Shuttle. He also helped develop a powerful video supercomputer for the Sarnoff Real-Time Corporation spin-off (which became DIVA), and helped found Sarnoff’s Pyramid Vision Technologies internal venture in 1997 to “productize” computer vision technology (U.S. 1, December 12, 1998). He also consulted for companies including Intel Corporation, Samsung, and Sensors Unlimited.
In 1999 Romanowich founded a new company, Home Animation, Inc. to provide wiring and media design and installation services to new home buyers — hooking up phone, networks, and audio/video, especially for home theaters (U.S. 1, January 12, 2000). But this was totally opposite from an engineering business — labor-intensive, and requiring a lot of hand-holding of new homeowners — so Romanowich eventually got out of the business.
Then in late 2001 Romanowich was asked back to serve as executive director of Sarnoff’s Pyramid Vision Technologies venture, which was developing smart video surveillance systems for customers including the military. In his two years on the job, Romanowich resolved yield problems by organizing the supply chain and helped triple sales to $5 million by using client input to improve products. The PVT video security products were eventually sold off by Sarnoff to L-3 Communications in 2004.
Through this period, Romanowich was looking for other business opportunities, and meeting other local technologists who could contribute to making them happen. He became involved in the Princeton entrepreneur community through organizations like the New Jersey Technology Council.
Romanowich also joined with Steven Georges, of Princeton Server Group, to pick up on the challenge of Rush Holt’s “Einstein’s Alley” initiative by founding a group to network with other entrepreneurs. This would directly benefit his new company. “We got out a lot of resources out of it,” he says, “including several employees.”
Romanowich had began focusing on the SightLogix concept with an informal circle of advisors in October, 2003, developing the business plan and formalizing the company with partners that include James Hahn and Eric Schwab.
“My previous ventures were basement operations,” he says, “and I realized I needed some real talent, so I reached out to my friend Jim Hahn.” Both Romanowich and Hahn were investors in the New Jersey Technology Council Venture Fund. Hahn brought experience from founding four start-up companies and was the first investor. He had taken a previous company, Infotron Systems, “from an actual basement to an IPO, with five international divisions and 1,400 employees.” As chairman of SightLogix, Hahn focuses full-time on the company.
Schwab, who had known Romanowich since NJIT in 1984, brought experience in product development, manufacturing, and outdoor packaging from positions at AT&T Bell Labs/Lucent.
The final founder was Danny Chin, who holds 28 patents in video processing. Romanowich had worked for Chin at Sarnoff’s SRTC (then DIVA) venture, where Chin was a cofounder and eventually director of advanced development. Romanowich says he invited Chin over to hear the company story, and “he joined the same night,” becoming vice president of engineering.
The company was formally founded as Automated Threat Detection (the predecessor company to SightLogix) in March 2004. Romanowich had been raising money from “visionary private individuals,” and the company received $250,000 in matching “Springboard” funding from the New Jersey Economic Development Authority for an initial seed funding round of $900,000.
“The requirements of the Springboard funding accelerated our fundraising,” says Romanowich. “We had a complete business plan already submitted to the EDA.”
As it became more established, SightLogix received an additional $750,000 in “Techniuum” funding from the EDA in July 2006. And it has taken advantage of the EDA tax transfer credit program, selling state tax losses to receive $240,000 from 2005 and 2006. “The EDA did a good job,” says Romanowich. “They have helped us grow our staff over 50 percent in the past six months.”
“New Jersey is smart,” says Romanowich. “They are putting their money where their mouth is, and they’re stepping up to make a difference on the world stage. This technology was developed in the state, and I was educated in the state, and there’s all the video-centric talent here that would be going away otherwise. Because of the resident skill base we know exactly where the best talent is, and how to attract it.”
Meanwhile, the technical challenge for SightLogix was to develop a complete camera system that would meet the needs of the market — not just security software, but the entire architecture, including the integrated camera and processing hardware. This required parallel paths of marketing and development while refining the concept. “We did a hundred interviews,” says Romanowich, “with end-users and market channels. It was too early for them to understand what we were trying to do, so we asked open-ended questions. We wanted to establish the underlining architecture, performance, and price points.”
The result of this technical and marketing work was a clear picture of the intended product, including a 100-page market requirements document and an engineering design document. Eight patents are pending.
In February, 2006, SightLogix expanded from incubator space on the Forrestal campus to its current location on Alexander Road. But Romanowich continued to avoid public notice. “We were like a skunk works with a technology that took a long time to develop,” he says. “We were running below the radar because the market was so chock full of false promises. We did not want to look like another company that was over promising and under delivering.”
Romanowich also expanded the staff. “John has been able to attract top talent from both commercial and government, technology and market,” says Hahn. “We have leading technologists and major government influencers on the team. It’s a strong team, with the company culture built around John.”
One such addition was Barfield, who joined as vice president of government programs. Barfield has a military and government background, doing high-tech security since 1990, and most recently running a test lab for the Marine Corps. “My role was to vet out cutting edge technology,” he says. “We had to see what was out in the market and bring them in and kick them around. We would actually install them, and bring them into the field for additional deployment. Some did not make it out of the lab. The ones that did were installed and had their own sets of problems.”
Barfield was introduced to Romanowich and SightLogix in February, 2006, by BAE Systems, a major government contractor and systems integrator that SightLogix had already been working for. “At that point I was ripe for generation three,” says Barield. It was very, very clear to me that what John and his team have put together here was light years ahead of what was available. So I wanted to actually go along for the ride.”
SightLogix needs to develop these connections because the system is not sold directly. Instead it is distributed though system integrators, who win the contracts for installation of security systems for major government and commercial sites. “They include our product and expertise into a larger security system,” says Romanowich, “which is validating our technology as something that has not been achieved to date.”
SightLogix has been selling its system for evaluation since June, 2006, and it already has been included in bids for major projects. Romanowich sees multiple opportunities that have the potential to install “a couple hundred” cameras over the next year, across multiple sites with four to five cameras each.
“Our product provides military level performance at commercial price points,” says Romanowich. “It has a global ubiquitous market for commercial and government applications for outdoor sites.” Romanowich says he has no competition, at least “at this level of sophistication.”
SightLogix began previewing and test marketing its system at security industry trade shows in 2005. At the ASIS International conference (originally the American Society for Industrial Security) in San Diego in September, 2005, says Romanowich, “people understood. They were surprised as well, saying ‘you solve the problems where current systems do not work.’”
Then by April, 2006, at the International Security Conference (ISC) West in Las Vegas, SightLogix was able to demonstrate a fully functional product prototype. “We now could take orders,”says Romanowich.
But it was back at ASIS International in September 2006 that SightLogix received a major coup, being chosen to exhibit in the Lockheed Martin booth.
“They selected only five vendors to be there,” says Barfield. “It’s a sought-after position — people look to Lockheed, and we were right smack in the middle of the booth.”
“Lockheed sees us as a leading technology of the future,” says Romanowich. “It brings credit to them as bringing the right innovations to their customer base.”
“Physical and cyber security have converged,” says Romanowich, “with global interest. It’s a $10 billion plus market. Companies like IBM and Cisco have jumped on the bandwagon, and we are well positioned to be working with these guys. This is the next major wave, and the magnitude dwarfs what has come before it. “
And SightLogix is positioned to catch that wave. Says Hahn: “We have sold product. It is installed and operating, and with happy customers — sites where other competitors were kicked out [for not performing]. And these are turning into major rollouts.”
“As we move more into the market the product is being very well accepted because of its architecture and its capabilities,” says Romanowich. “Everybody who sees it falls in love with it. This is especially true of the people who really understand it — we presented it to somebody today, and before we finished they said they wanted to invest. The product makes so much sense in the way that it does things, and the economic proposition is so strong.”
SightLogix, 745 Alexander Road, Suites 5 and 6, Princeton 08540; John Romanowich, president. 609-951-0008; fax, 609-951-0024. www. sightlogix.com
See Doug Dixon’s www.manifest-tech.com for commentary on multimedia technology.