Ken LaMarca, CEO of Active Intelligence outlines the difference between how anomaly detection learns behaviors versus looking for specific events and actions.

Benefits of video security

Recently, the industry has been buzzing about video analytics technologies and how they have the intelligence to help prevent potentially threatening events from escalating and even proactively detect threats before they take place.

This is undoubtedly a game changer for physical security and law enforcement professionals looking for more effective ways to reduce crime and enhance overall security. Video analytics also provide security professionals with the long-awaited solution they need to substantiate the budget for more advanced security technologies.

This has always been a challenge without demonstrating how such investments can deliver real return on investment (ROI). Given the wide range of security and business intelligence applications across the enterprise and relative ease of deployment for most solutions, video analytics can easily be cost-justified.

The significance of anomaly detection

However, legacy video analytics technologies have a fundamental limitation: they’re rules-based and only detect what they are programmed to see, requiring the use of application-specific software modules.

This can be costly to implement and inefficient in detecting events that require action. For example, if you employ an analytic solution that autonomously detects weapons and a fight breaks out, the latter will go undetected. So, unless the video system is being monitored by live personnel who spot the weapon and take immediate action, the event will be overlooked until reported – at which point the situation may be serious or deadly.

All this has changed with an emerging technology that far surpasses the capabilities of traditional video analytics – anomaly detection. Anomaly detection employs machine learning algorithms that constantly adapt to events in a surveilled area. If something as simple as a person falling or lying on the ground occurs in an area where this type of event is statistically uncommon, the solution detects the event in real-time and alerts designated personnel or triggers a Standard Operating Procedure (SOP).

This applies to any irregular activity that takes place in the designated camera’s field of view – fire and smoke, fighting, weapons brandishing, irregular movement of vehicles or people, vandalism… anything that is simply not supposed to be taking place in a given area.

Anomaly detection software can be selectively applied to any single camera, group of cameras or system wide via integration with a video management system (VMS). This allows users to better leverage their existing investment in video systems, transforming these systems from reactive to proactive tools. Artificial intelligence (AI)-powered software can tend to have hefty hardware requirements, but lightweight options with modest hardware needs do exist. Choosing an option that doesn’t require expensive high-performance hardware can ease the cost justification process.

Implementation of these solutions is often simple. Once installed on a video appliance server, or embedded into a VMS, the anomaly detection solution spends a period of time monitoring the selected video feeds, using AI and machine learning to build a baseline of what it would consider “normal” for each individual camera’s field of view.

Once the learning period is complete, the anomaly detection software notifies authorized personnel in real-time when an abnormal activity is detected. This allows security personnel to quickly take action to remediate the situation, if necessary. The proactive nature of the solution can help prevent potentially dangerous events from escalating and/or help alert personnel if an individual needs assistance.

Uses for anomaly detection

Unlike conventional rules-based video analytics, anomaly detection software has virtually unlimited applications for physical security, health and occupational safety and business intelligence. Anomaly detection software can automatically call attention to a virtually endless number of event types.

Examples include but are not limited to: irregular movements of people and vehicles; medical emergencies; suspicious behavior; vandalism and destruction of property; pooling water and flooding; dangerous vehicle movements; slip and fall incidents; theft and home invasion; workplace accidents; crowd gathering and dispersion; unscheduled visitors and trespassing; fire and smoke; and aggressive behavior and fighting… essentially, anything out of the norm will be detected.

Here’s a perfect scenario of an abnormal event that conventional video analytics simply wouldn’t notice. Employees are returning to the workplace, but often on flex hours or hybrid schedules. Some employees enjoy working at odd hours or on the weekends when offices are generally not occupied. An employee falls or is stricken with some other medical ailment and is lying unconscious on the ground.

Anomaly detection software would automatically report this unusual event, as it would individuals ransacking an office or retail space. Any action outside of “normal” activity for this specific area under surveillance in this specific office environment would trigger an alarm in real-time. An added benefit of anomaly detection is that it never stops learning and will continue to learn new events and behaviors over time including season changes, traffic flow, weather, schedule changes and more.

Anomaly detection is also completely unbiased, without profiling or infringement on personal privacy. No images or personal information is ever recorded and stored for reference. All anomalies are identified based on statistical data captured by the software without any human judgment. This makes anomaly detection ideal for use in nearly any public and private environment.

Choosing video technology

Anomaly detection technology sets a new benchmark in preemptive video technology for a myriad of physical security and business intelligence applications across organizations of virtually any size. Unlike most new technologies, anomaly detection is easy to cost-justify given the vast numbers of ways it can mitigate risks and liabilities.

Plus, it’s fast and easy to implement using existing video systems, does not require special training, effectively reduces the need for live system monitoring, increases security and safety and minimizes risks and liabilities. Anomaly detection software may turn out to be the picture-perfect solution the industry has been waiting for, offering preventive and proactive video security and business intelligence applications.

Active Intelligence is a video security technology company focused on transforming video security systems from a forensic resource to a prevention tool. Started in 2019 by a trio of security industry leaders, Active Intelligence has a combined 100+ years of experience in corporate security, public safety and homeland security.

This article was originally published in the February edition of Security Journal Americas. To read your FREE digital edition, click here.