Avigilon and Video Analytics

 

Avigilon is a Vancouver based security equipment designer and manufacturer, perhaps most well known for the software they have developed to analyze raw video data, with no input other than the pixels streamed through an HD camera setup. “Open Video Management” using what they term “Self-Learning” video analytics is the cornerstone of any modern video analytics solution. The company also manufactures several lines of high quality HD cameras and access control gear, but it is not a requirement that all equipment be from Avigilon. Avigilon analytics is performed on-site with a proprietary set of hardware and software, and is effectively limited only by the quality of the video input.

What does this all mean for the consumer? For the residential market, there is not much application, and most home owners would see the cost as extraordinary, considering the increasingly cheap residential offerings from mainstream security integrators. It is also generally not necessary to run such extensive analytics in a residential setting. What Avigilon’s video analytics are particularly useful for are larger, enterprise-scale security operations, such as car lots, schools, and highly trafficked gates. This is something our company has considerable experience with, both in installing and integrating with other cameras and access control systems, as well as servicing and maintaining this type of equipment.

For example, let’s look at our most popular security solution featuring Avigilon; car dealerships. Avigilon’s analytics recognize a car versus some other object and distinguishes between a person and an animal or some other moving object which is not a security threat. The degree to which any software is actually “self-learning” is certainly up for debate, though the type of software underpinning these analytics systems is similar in structure to the kind of predictive software used by Google, Amazon, or Facebook to predict future purchases and analyze behaviors. The idea is that the software is able to self-code within the limited scope of object detection and threat discernment. This implies that each individual system that is integrated with Avigilon analytics will be slightly different over time for each installation in order to better perform in that particular setting.

Security Technology of South Texas is well versed in customizing analytics solutions to each customer based on their concerns and security needs. As we move into the future, analytics will become a de facto component of any competent access control and video surveillance installation, as it is able to reduce or eliminate false alarms, the traditional bane of functional, digital security.

Avigilon has recently announced what they call a “next generation” AI capable camera system. Called the “H5 Smart Camera”, this technology makes use of Deep Neural Networks built into the devices themselves to provide the ability to learn, detect, and notify for events requiring investigation. So called “deep learning” or “deep neural networks” are the same type of AI tech used in the AI operations of companies like Google and Amazon. Deep learning AI works by sorting through massive amounts of data, the more the better, and modifying its own code based on what the program sees. Google’s “Watson” is a famous example of this kind of software, and gained public exposure when it defeated the world’s top player of “Jeopardy”. Deep learning was also used in the software that established computer’s as the top “players” of Go, a chess-like strategy game which is well known for its highly technical play and difficulty.

Avigilon is working with Intel, a leader in the emerging AI marketplace. Intel’s “Movidius” VPUs allow for power-efficient acceleration of hardware in deep neural networks. These networks are inspired by and in some ways modeled from the way that biological systems (brains and central nervous systems) process and interpret information. Intel believes this technology will “usher in a new generation of video analytics capabilities with deep learning approaches” (avigilon.com). The H5 camera line uses these AI facilities to develop more sophisticated event detection and automatic notifications. The cameras are to be displayed at the Global Security Exchange conference of 2018 in Las Vegas, and are expected to be launched into the market in 2019.

This is the focus that Avigilon has had in the industry for a while already and they have analytics systems available now. Yet as technology marches up an accelerating curve in processing power, we can expect more and more functionality in all AI systems to come. AI itself is a fascinating technology that we have looked at in more detail in previous articles, and is expected to radically change the markets in almost every industry and the global economy as a whole.
Even though huge volumes of video data are collected every day, most statistics indicate that only 10 percent of this data is ever used. The majority of data collected loses its value very quickly after being generated. The reason for this? Our primary focus tends to be delivering the correct information in a crisis or providing it as evidence after criminal activity has taken place. This causes much data to be “wasted” in the sense that we lose our on the opportunity to perform useful analytics.

Video analytics is an increasingly powerful tool. It helps to improve usability of these vast amounts of video information. Analytics software acts as the “brain” of a surveillance system and is built into IP cameras themselves or processed in separate computing infrastructure. This creates a smarter system that “knows” what it sees and alerts guards to potential threats as soon as an alarm rule or condition is met. Analytics gives operators the chance to act faster and more efficiently with better intel.

Video analytics is like an ever-vigilant system operator within the security system itself. It captures data like a panopticon, seeing all in every monitored scene around the clock. Content analysis information, a form of video metadata, is stored as well. As they reduce operating costs and increase efficiency, intelligent cameras deliver a solid return on investment which can be measured in tangible results to the business or other setting in which it operates.

Let’s take a look at what exactly is possible using intelligent video. Smart IP cameras are able to classify the objects they see on their own. Objects entering or leaving the scene can be identified as a person, car, bike, truck, or other object. Because the camera can differentiate objects, it can be told to only alert in the case of a break-in, ignoring things like leaves in the wind or animals wondering through. New low-light cameras allow color-filtering even in scenes with very little ambient light. Even at night, color detection is possible in this way.

Alarm detection can be set to be even more specialized. Rules can be configured to look for specialized behavior patterns such as fighting, running, loitering, path following, abandoned luggage, entering an area, and more. The alarm engine in each camera coordinates with the others in a logical way to interpret this information and determine threat status. All this allows for a very robust alarm condition solution and prevents false and missed alarms.
What’s more, stored metadata enables forensic analysis at a later time. This means that retroactive searches for a specific car or person is possible even if it was not a determined item of interest until well after the event was recorded. Metadata is compact and only barely adds to the size of recordings. It is quick and easy to search through to find a specific event.

The logical next step is to continue to improve analytics for video metadata until we approach 100 percent practical use. Predictive analysis of human traffic patterns can predict shoplifting and identify potential events before they take place, and the more data that can be made useful the more accurate these types of predictions will be. The same technology can monitor customer dwell time at different displays in a store and determine the effectiveness of in-store advertising and product locations. Analyzing customer engagement with these displays can help increase customer engagement with products and lead to increased sales and revenue. As the IoT expands, this type of technology will be more and more critical as there will be many more points of data to analyze. There is no way to fully anticipate the eventual effects this will have on our industry or the world at large.

Sources: Avigilon.com

—————————————————————————————————————————————–

Part 2: Facial Analytics

The American Civil Liberties Union recently tested Amazon’s facial recognition tech — and the results were less than favorable. To test the system’s accuracy, the faces of all 535 members of congress were scanned against 25,000 public mugshots, through Amazon’s open Rekognition API. Although none of the members of Congress were in any of these mugshot lineup, Amazon’s system nevertheless generated 28 false matches. The ACLU claims this raises some particularly serious concerns about Rekognition’s use by law enforcement and in the legal and medical world.

“An identification — whether accurate or not — could cost people their freedom or even their lives,” the group said in an accompanying statement. “Congress must take these threats seriously, hit the brakes, and enact a moratorium on law enforcement use of face recognition.” (ACLU)

According to The Verge, an “Amazon spokesperson attributed the results to poor calibration.” However this does not necessarily account for the results. Amazon’s system currently operates with the default confidence threshold of just 80 percent. Yet Amazon claims it recommends at the very least a 95 percent threshold for situations such as medicine and law enforcement where relying on a machine to ID someone could cost them their freedom, life, or worse.

“While 80% confidence is an acceptable threshold for photos of hot dogs, chairs, animals, or other social media use cases,” the representative said, “it wouldn’t be appropriate for identifying individuals with a reasonable level of certainty.” (ACLU) Even still, the Rekognition suite does nothing to affect that recommendation during the process of setting it up, and there is of course little to nothing to prevent law enforcement agencies from using the default setting of 80 percent.

In May of this year, this tech came into the limelight when the ACLU report was able to show the system being in use by a number of LEO agencies including the police of Orlando, Florida. It is sold as a part of Amazon’s Web Services cloud, and is quite inexpensive with a costs as low as less than just 12 dollars a month for the entire department.

Furthermore, this test demonstrated a continuing problem of many facial recognition systems, which have  historically had considerably difficulty    in accurately identifying both women and non-white minorities. Of the 28 false matches, 11 involved black members of congress, although they make up just around 20  percent of the whole of congress itself. Some other systems fair even worse. With the system used by the London Metro Police force producing as many as 49 false matches for every legitimate hit, which then necessitates a manual and time and resource consuming search though these false-positives.

Ostensibly, facial recognition IDs would be confirmed through multiple human sources before an arrest would be made, though many say that even checking faces violates privacy rights. Worse still, it is not hard to imagine a situation where an officer sees a false match that leads him to believe the potential arrestee could be armed and dangerous, and also plant ideas about the person before even really investigating, changing the outcome of a routine stop from routine, to possibly violent, even deadly.

Security Technology of South Texas works with analytics and facial recognition video surveillance in its projects, and are experts in integrating, understanding, and sourcing only the best tech to get your job done, at a price you can feel good about. Let us show you the difference between a local, responsive, company that strives for only excellence and client satisfaction versus the kind of experience we have all come to expect from the detached, hard to reach, and inferior service and installations inherent to the juggernauts of the security industry.

Security Technology of South Texas is an authorized integrator for Avigilon systems and has designed systems with this kind of functionality. Avigilon analytics is particularly useful in enterprise scale operations, school and college campuses, as well as car dealerships or any other large property where tight security is necessary.

Please contact us at  admin@gostst.com on our website

or via phone at  210-446-4863   24/7.

Sources: ACLcomU, Verge.com,  Amazon.