You’ve likely seen us use “Anomaly Detection” to describe our flagship product, ASTRA. While it may be obvious from the name that the technology identifies anomalies in video feeds, how it works and why it’s different from everything else on the market may need some more explanation. In this installment of ASTRAL INSIGHTS, we’ll dive into how anomaly detection works, how it’s different, and how it can be used.
What is anomaly detection?
Anomaly detection is technology our in-house mathematicians and software developers have created to detect outliers in video feeds in real-time. When applied to an existing video security system, it puts virtual eyes on every video feed, allowing for 24/7/365 real-time monitoring, without the need for operators to view every camera stream.
When an anomaly or outlier occurs in the video feed, ASTRA can be configured to notify operators, trigger a 3rd party software, send an email or SMS message, or push to a Mobile Client. This notice will allow operators to bookmark or save footage, and intervene if necessary. How your team reacts to these notifications is up to you and your standard operating procedures.
How does anomaly detection work?
The power of ASTRA’s anomaly detection is founded on its ability to constantly learn and adapt. It begins to learn about your facility and video from the first day it is activated on your system. It takes approximately two weeks to build a model for what the system would consider “normal”.
From here, ASTRA’s anomaly detection engine, powered by artificial intelligence and machine learning, compares a series of video frames to its model of normalcy to determine if an anomaly has occurred. If enough deviation from normal has occurred, operators are alerted in real-time.
However, the learning process doesn’t stop after two weeks – what is normal today, may not be normal in a month or a year. ASTRA’s detection engine continues to learn every single day. It takes into account minor environmental changes and gradual shifts within the video feeds to ensure that it is alerting on events of interest, and not (for example) snowfall or general foot traffic changes.
Why choose anomaly detection over traditional video analytics?
You don’t know what you don’t know
Traditional rules-based video analytics systems can be configured to alert operators to very specific events. While this can be useful in some instances, these rules-based solutions will only alert for situations that they are configured and licensed for. Most security teams can safely assume that they’d like a notification for perimeter crossings or an oddity like a vehicle in a pedestrian walkway, but there are literally infinite situations that cannot be planned for.
Rules-based analytics require that you can predict the future. If that was possible, why would anyone bother with a video security system?
ASTRA’s Anomaly detection does not function within the confines of a rules database. Instead, it looks for any situation that is outside of the norm – no rule configuration or programming required. Most importantly, anomaly detection can identify events of interest that you and your team hadn’t planned for – often some of the most devastating.
Cost
Rules-based systems are often licensed in a way that requires users to purchase modules for each event they’d like to be notified about. Each of these modules requires additional computing power, which directly translates to added cost in the form of computing hardware, electricity, physical space, HVAC requirements, maintenance, and more.
Anomaly detection like ASTRA’s doesn’t require additional module purchases to alert on specific events and instead alerts on anything abnormal in the video scene. Anomaly detection solutions are designed to be lightweight from the outset and can process up to 10 times the number of camera streams using modest, off-the-shelf hardware. The software is so light, that it can even be run on the edge, a feature of ASTRA that is currently in development.
The video security industry is pushing ahead and leaving antiquated rules-based analytics behind. Anomaly detection is the way of the future for video security, and it brings with it cost savings, better and more accurate alerting, and – most importantly – increased safety and security.
Interested in learning more about how ASTRA’s anomaly detection works? Contact our sales team to schedule a demo today.
Subscribe to ASTRAL INSIGHTS to get updates when new blogs are posted.