What is AISight?
AISight: The World’s Only Behavioral Recognition™ System
AISight®, created by BRS Labs, is the ONLY video surveillance software that meets the needs of today’s ever-changing security environment.
The potential for security threats has grown dramatically over the past decade. Despite the advances made in other areas of security infrastructure, video surveillance technology’s evolution did not meet the needs of the market. To address these more sophisticated threats, BRS Labs developed AISight — a revolutionary product that has changed the security industry forever.
Traditional video analytic software can only compare captured video activity to a list of preprogrammed objects and scenarios. It requires the costly setup of tripwires, zones of interest, and scene boundaries. Initial setup and subsequent maintenance is labor intensive and has demonstrated a poor return on investment.
BRS Labs’ AISight is advanced, intelligent software that uses Behavioral Recognition™ technology to learn—on its own—about the environment and objects it observes in each camera’s field of view. Since its learning is perpetual, AISight understands which activities commonly occur in any particular scene, bringing attention to objects or behaviors that are out of the ordinary through real-time alert notification. It begins autonomously learning about every environment it observes from the moment it is connected to a video network … AISight never requires the burdensome preprograming (or reprograming) necessary for legacy video analytic systems.
With eleven Registered Patents granted and more than fifty others pending, AISight is a unique, ground-breaking product whose capabilities surpass all other video analytic products. It provides accurate, real-time alerts to security personnel about real threats, while constantly learning to ignore the everyday behaviors that trigger the exponential number of false alarms in other products. Busy security personnel no longer need to waste valuable time with systems that continually cry wolf.
How AISight Works
AISight works with patented learning and analysis engines that enable the system to observe events, analyze them, and remember them similarly to how human brain makes and stores memories. When new events differ from AISight’s memories, it determines that a suspect event is occurring and alerts security personnel.
After the software has been started, it connects to the video network and begins to monitor the unique environment and activities for each individual camera. Each camera view is stored as a separate memory. Elements that are always present in the environment become part of the “background.” Objects that enter the field of view are analyzed based on their appearance, classification and interaction within its environment and other objects. AISight analyzes the structures, sizes, shapes, locations, velocities, accelerations, paths of objects and other characteristics of all objects within the scene and forms memories about them. It also records timestamps for these events and remembers during what times of day or days of the week events most frequently occur. Just like the “Long-term Memory” of the Human Brain, the more frequently certain objects and behaviors are observed, the stronger those memories become.
Whenever AISight observes objects and behaviors, it compares these events to its current memories. The less frequently it has observed an event in the past, the weaker its memory will be about the event and the more unusual it will deem the current activity. Unusual activity is immediately reported to security personnel to enable a proactive response to potential threats, but normal activity is ignored. And even when AISight has learned to ignore certain activities, it can still be told to alert security personnel of those activities regardless of how often they occur, if needed.
Just as frequent observation of objects and events reinforces AISight’s memories, memories that aren’t reinforced degrade. This means that AISight not only learns about commonly occurring activity but also “forgets” when that activity becomes less frequent, enabling it to alert on events that are no longer commonplace. Because of this unique ability to learn, remember, and forget, AISight’s ability to provide currently relevant, accurate alerts evolves alongside the environment. It adapts to moving vegetation, lighting changes, repositioning of furniture, weather patterns, and myriad of other environmental aspects that challenge video analytic systems.
What is Behavioral Recognition Technology?
Traditional video analytic software can only compare captured video activity to a list of preprogrammed objects and scenarios. It requires the costly setup of tripwires, zones of interest, and scene boundaries. Initial setup and subsequent maintenance is labor intensive and has demonstrated a poor return on investment.
BRS Labs’ AISight is advanced, intelligent software that uses Behavioral Recognition™ technology to learn—on its own—about the environment and objects it observes in each camera’s field of view. Since its learning is perpetual, AISight understands which activities commonly occur in any particular scene, bringing attention to objects or behaviors that are out of the ordinary through real-time alert notification. It begins autonomously learning about every environment it observes from the moment it is connected to a video network … AISight never requires the burdensome preprograming (or reprograming) necessary for legacy video analytic systems.
With eleven Registered Patents granted and more than fifty others pending, AISight is a unique, ground-breaking product whose capabilities surpass all other video analytic products. It provides accurate, real-time alerts to security personnel about real threats, while constantly learning to ignore the everyday behaviors that trigger the exponential number of false alarms in other products. Busy security personnel no longer need to waste valuable time with systems that continually cry wolf.
How AISight Works
AISight works with patented learning and analysis engines that enable the system to observe events, analyze them, and remember them similarly to how human brain makes and stores memories. When new events differ from AISight’s memories, it determines that a suspect event is occurring and alerts security personnel.
After the software has been started, it connects to the video network and begins to monitor the unique environment and activities for each individual camera. Each camera view is stored as a separate memory. Elements that are always present in the environment become part of the “background.” Objects that enter the field of view are analyzed based on their appearance, classification and interaction within its environment and other objects. AISight analyzes the structures, sizes, shapes, locations, velocities, accelerations, paths of objects and other characteristics of all objects within the scene and forms memories about them. It also records timestamps for these events and remembers during what times of day or days of the week events most frequently occur. Just like the “Long-term Memory” of the Human Brain, the more frequently certain objects and behaviors are observed, the stronger those memories become.
Whenever AISight observes objects and behaviors, it compares these events to its current memories. The less frequently it has observed an event in the past, the weaker its memory will be about the event and the more unusual it will deem the current activity. Unusual activity is immediately reported to security personnel to enable a proactive response to potential threats, but normal activity is ignored. And even when AISight has learned to ignore certain activities, it can still be told to alert security personnel of those activities regardless of how often they occur, if needed.
Just as frequent observation of objects and events reinforces AISight’s memories, memories that aren’t reinforced degrade. This means that AISight not only learns about commonly occurring activity but also “forgets” when that activity becomes less frequent, enabling it to alert on events that are no longer commonplace. Because of this unique ability to learn, remember, and forget, AISight’s ability to provide currently relevant, accurate alerts evolves alongside the environment. It adapts to moving vegetation, lighting changes, repositioning of furniture, weather patterns, and myriad of other environmental aspects that challenge video analytic systems.
What is Behavioral Recognition Technology?
Behavioral Recognition™ is a Revolutionary Technology Which Fundamentally Changes the Video Surveillance Market
Behavioral Recognition technology, the backbone of the AISight system, combines computer vision with machine learning to provide actionable intelligence through real-time, relevant alerting of anomalous behavior observed by cameras. It is fairly simple to program a computer to detect movement with a camera. It’s just as simple to trigger a response if that movement violates a condition — a “rule”. But the real world environment is anything but simple. The slightest ambient variation, such as a shadow, can wreak havoc on these preconceived rules and the result is, as it has been for the last ten years, unmet expectations, disappointment and frustration by users.
BRS Labs has patented the process and technology now known as Behavioral Recognition. This technology was invented to achieve the effect without the cause, something the rules-based systems on the market today have been unable to deliver. The real world is not black and white, on or off, yes or no. The rigidity of a simple rule will always break. “Behavior” is not black or white either; it’s an endless amount of colors hidden by an infinite amount grades and tints. The trick is to recognize behavior (the cause) and achieving the effect (the distinction of behavior).
The number of shades of the color red is endless and with so many variations a human cannot explain the difference he sees between the most subtle shades. But yet one still knows when two articles of red clothing don’t look right together, that they don’t match; this is the concept of Behavioral Recognition. Introduce dimension and you have a different example: How does one teach a computer to recognize a human when seen through a camera, while every camera sees from a different angle? One can’t! A single human will look completely different from one camera to another. One camera mounted low may be able to detect four limbs and a head, while a camera mounted high might not even see the legs as a person walks below it. The only viable method is to allow the computer to learn the subtle differences itself. Let it become aware that objects appear differently from one angle to the next, and that the behavior of humans in one camera may be completely different than the behavior found in another. Only then can the power of reason be utilized. Behavioral Recognition is a “Reason-based” system unencumbered by man-made rules.
First utilized in San Francisco's MUNI system.
A new breed of security cameras can supposedly detect terrorism and crime without a human judgment call--and mass transit agencies are shelling out big bucks for the product. San Francisco's Municipal Transit Authority, which oversees the city's MUNI trains, has signed a contract with security firm BRS Labs to deploy cameras to 12 subway stations that use algorithms and machine learning techniques to spot anomalous behavior.
BRS Labs is a security firm that provides behavior recognition software for video surveillance. The company's clients include government, tourist attractions, military bases, and private industry; BRS's software issues real-time text alerts when cameras detect strange behavior. Servers connected to security cameras observe locations for weeks at a time and then establish a baseline of “normal” behavior based on this timespan; anomalous activities afterwards (loitering, abandoned packages, abnormally high/low numbers of passengers) trigger an alert. No tripwires or programming of initial parameters are required.
According to a publicly available product bid, the San Francisco MUNI project will include up to 22 connected cameras at each train station; video monitoring will be conducted by train control, MUNI Metro East facility, and in-station personnel. The video systems will build memories of observed behavior patterns that mature with time; the systems, in the bid's words, “[have] the capability to learn from what [they] observe.”
In an interview with Fast Company, BRS Labs President John Frazzini said that the company's AIsight behavior recognition product relies on 11 patents related to computer vision technology and surveillance imagery. BRS's patents primarily deal with the intersection between computer vision and machine learning; video footage grabbed by MUNI cameras will be automatically translated into code for real-time processing. Clips of anomalous activity are dispatched to MUNI employees automatically; SMS text message alerts are also sent to staffers' mobile phones.
The post-9/11 emphasis on “homeland security” and anti-terrorism efforts has resulted in a gold rush of surveillance contracts from mass transit agencies and public institutions nationwide. While large mass transit agencies such as New York's MTA and Chicago's CTA have been cagey about their counter-terrorism efforts, trade show presentations and chatter in industry publications have given a basic idea of what is happening. Apart from machine learning-based video surveillance, subway security has also taken on wackier (and scarier) aspects: The Homeland Security Department has publicly announced their plans to release bacteria into Boston T tunnels this summer in order to test new biological weapon detectors deployed throughout the subway system.
The same technology that's being deployed in San Francisco's subway is also intended for the global market. BRS, which is based in Houston, has overseas offices in London, Sao Paulo, and Barcelona. BRS Labs' AISight product is primarily intended for use in counter-terrorism efforts. AISight's software algorithm has limited success in detecting in-station muggings or subway perverts, two issues of much more immediate interest to mass transit ridership than terrorist attacks.
Another unique aspect of BRS's product is the fact that it heavily relies on timestamps and time recognition. Behavior and objects are coded according to the times of day or days of the week in which they most frequently occur; the velocity, acceleration, and path of customers passing through a station are analyzed as well. Spatial anomalies and classification anomalies are taken into account as well.
One worrying--or appealing to budget-minded clients--aspect of BRS's product is the fact that their software product sharply reduces the need for human camera maintenance. The algorithms behind AISight compensate for lighting changes, shaky images, and poor bandwidth. Between the automated evaluation of “anomalies” and their software-based maintenance process, the need for human supervision for effective software operation sharply declines.
BRS's promotional literature promises that their software product can accurately detect loitering in unusual places at train stations, abandoned objects, and “tailgating” at entrances.
Verified customers of BRS's system beyond the SFMTA include the City of Houston, Boeing, the Louisiana Port Commission, the City of Birmingham (AL), and security contractors for the Nuclear Regulatory Commission. Publicly available documents indicate that the Port Authority of New York and New Jersey is deploying BRS's technology for a pilot project at the World Trade Center as well. Fast Company is based at the World Trade Center complex.
0 comments