Video Analytics and Avigilon Appearance Search
“It’s all about the analytics,”
states Gary Sherry.
As regional sales manager in the San Francisco Bay Area for Avigilon.com, a Motorola Solutions company, Sherry works for one of the premier video analytics companies in the world.
The Vancouver, B.C., Canada-based company has designed, developed and manufactured advanced security surveillance solutions, including video analytics, network video management (VMS) software and hardware, surveillance cameras and access control solutions for nearly 15 years. The innovative company’s products are used by a range of commercial and government customers including critical infrastructure, airports, government facilities, public venues, healthcare facilities and retail centers. Not only that but Avigilon holds more than 750 U.S. and international patents.
Factor in that Motorola Solutions acquired Avigilon for $1 billion in the first quarter of 2018, and an even brighter future emerges for the provider of end-to-end surveillance solutions. When Avigilon Founder Alexander Fernandes announced the sale, he noted the acquisition would bring new opportunities to Avigilon, enabling it to accelerate its innovation and provide even more value to customers.
Two new developments show Fernandes’ predictions are bearing out. In July 2018, Avigilon received SAFETY Act Designations from the U.S. Department of Homeland Security for its video surveillance systems and Access Control Manager (ACM) software. In March, the firm announced it will preview the newest version of its video management software, Avigilon Control Center (ACC) 7 at ISC West 2019.
ACC 7 introduces Focus of Attention, a cutting-edge user interface for live video monitoring. Rather than watching static video feeds, Focus of Attention uses artificial intelligence-powered Avigilon self-learning video and analytics and unusual motion detection technologies to determine what information is important and should be presented to security operations.
Sherry reports this software tool replaces reactive surveillance methods with proactive ones. In the past, operators installed ip cameras, and no one monitored them until something happened. Surveillance shifted again when companies began putting a human being in place to monitor security cameras, which was better but not perfect, he says. He explains, “You can have someone staring at a monitor all day long, but unless something really jumps out at them and comes to the forefront, events can happen and they don’t know they even happened. Our goal now is to be more proactive and to stop bad things from happening. We can be proactive by using a security camera analytics software.”
Now when Avigilon security cameras spot people or vehicles that should not be in a surveilled area, a thumbnail of the scene pops up on operators’ screens with red boxes drawn around the suspicious activity. This calls attention to the situation so operators can act. They might talk into a microphone or flash lights to scare the person off or they might call the police. The point is that they can act because the situation has been called to their attention automatically.
Avigilon ACC 7 Software
Whether ACC 6 or 7, Avigilon’s VMS gives operators full control over their workstations with a simplified screen layout, intuitive controls and customizable features that improve multi-person interactions.
The software leverages advanced artificial intelligence and video analytics, including Appearance Search technology and Unusual Motion Detection technology, and machine learning to help operators quickly detect, verify and act on events. The bookmark, event, alarm and thumbnail search options in the VMS enhances the way operators interact with their systems, dramatically improves event response times, and helps make video actionable.
License Plate Recognition
ACC also employs Avigilon License Plate Recognition (LPR) analytics, which automatically reads license plate information from vehicles and links it to live and recorded video.
This enables security operators to search and quickly find specific captured license plate video for verification and investigation. This tool captures, recognizes and searches license plates with high accuracy. Operators can even create and import license plate watch lists so the system can provide security alerts when it spots specific license plates.
In addition, the VMS also features a loitering detection capability that triggers a video analytics event when an object or person moves into a region and stays for a fixed period. The operator can then act to get the person or vehicle moving along or summon police or security to the area.
WHAT IS APPEARANCE SEARCH TECHNOLOGY?
The new version of ACC, as well as previous versions, leverages Avigilon Appearance Search video analytics technology, which is a sophisticated deep-learning, artificial intelligence search engine for video. It sorts through hours of video with ease to quickly locate a specific person or vehicle of interest across an entire site, or multiple sites that are connected to the same ACC.
Operators can use this tool to initiate searches based on physical descriptions, including clothing, color and gender. If they incorporate the unique characteristics of a person’s face into the search, it enables Avigilon Appearance Search technology to understand that it is searching for the same person in the scene, even if items such as the person’s clothing changes over time. Operators can also quickly comb through hours of recorded footage, efficiently grouping video data, to help track a vehicle or individual’s route, identify a previous and last-known location and assist investigations. Playback, bookmark and export tools help the user build a comprehensive set of video evidence from multiple video sources, to create a powerful narrative of events.
Sherry reports two types of searches are possible with Avigilon Appearance Search. When operators are looking for a person of interest, they can right click on the person in video footage and ask the system to retrieve all video with that person in it. When operators do not know anything about a person’s facial features, but know what he was wearing, they can instruct the software to look for specific characteristics, such as a man with brown hair, a plaid shirt and blue jeans. The software then retrieves thumbnails of individuals matching this criterion, and operators can sift through them till they find the right person. They can then ask the system to surveillance footage for that individual.
“The software does what is called a thumbnail search,” Sherry says. “Within about four to five clicks, and under two minutes, you can find exactly what you are looking for.”
He explains the system learns and trains itself when a query is made. “If there is a suspected shoplifter at a mall, for instance, the system can locate him getting out of the car, track every place he went, and his return to his car, to build a storyboard. As he leaves, a different camera can capture their license plate number,” he says.
With that information in hand, the operator can search video footage from the last 30 to 90 days to learn when that person comes to the mall and what he does when he arrives. With his license plate number, operators can ask the system to inform them every time he returns. “Now security knows they are coming,” he says.
Installation and Setup of Avigilon
To leverage Appearance Search technology, a security camera installation company selects Avigilon IP security cameras with self-learning video analytics and combine them with Avigilon NVRs that are pre-loaded and pre-configured with ACC software. This enables Avigilon Appearance Search technology site-wide.
“We make everything,” says Sherry. “We have the largest selection of cameras in the world. We make the VMS software, which we call ACC. We make hardened appliances that everything runs on, and we do all the analytics on the camera, not on the server. And, every one of our analytic cameras, out of the box, already knows what a million people and half a million vehicles look like.”
After a security company installs an analytic camera, it immediately enters learning mode to learn the scene. From there, operators can add things they do not want the system to worry about. For instance, if operators do not want surveillance cameras to alert on maintenance golf carts, they tell the system a golf cart is not a vehicle, and from then on, the software ignores golf carts.
Because the system performs camera-based analytics, Sherry reports it eliminates the need to send all meta data to a central network for processing. It is done within the camera, which can process video faster with less overhead. There is no need for a powerful server in the core to process the data, Sherry says.
Installing analytics security cameras can increase the costs of a project, making it more cost efficient in many cases to only install them in key areas. Sherry reports it is often unnecessary to have analytic security cameras in an entire area being surveilled. “You don’t need your security company to install an analytics camera to monitor a supply closet,” he explains. “But you do want analytics cameras in your parking lots, your loading docks, your lobbies and your entrances.”
PROPRIETARY HARDWARE VERSUS OPEN SOURCE
While there are advantages to using a single vendor for an entire system, Avigilon understands that is not always possible. For this reason, if a client needs a new camera system, but already has some of the other pieces in place, they can pick and choose from what Avigilon has to offer.
Sherry states, “At Avigilon we write one driver and one driver only, and that’s an analytic driver.” This allows operators to use Avigilon cameras on other systems and vice versa. If someone just purchased a new surveillance camera system but hates its VMS, Sherry says, “We can go in and replace the VMS with ACC and roll all of their cameras into our system.”