When a security team monitors dozens or hundreds of cameras across multiple locations, recognising genuine threats from harmless activity can be tricky.

Without filtering alerts properly, Alarm Receiving Centres (ARCs) can become overwhelmed. Operators can experience fatigue and responses may be delayed. Or worse – genuine threats can get lost among the never-ending stream of alarms triggered by things as benign as swaying branches.

Of course, managing false alarms is necessary for ARCs that want to be accredited by bodies like the National Security Inspectorate (NSI).

Thankfully, AI event filtering – based on video analytics – automatically filters out non-threatening events, helping operators focus on the alerts that matter. In fact, the video analytics market is forecast to reach USD 22.6 billion by 2028, up from USD 8.3 billion in 2023 (a CAGR of 22.3%).

This article explains how it works, its many applications, and how it can help you deliver more robust services than ever.

What is AI Event Filtering?

Alarm monitoring software for central station automation steps up a notch with AI event filtering. This technology uses artificial intelligence to analyse video feeds and determines which events are genuine security threats. Only genuine threats are flagged for human operators to review.

Traditional motion detection systems can trigger alarms based on any movement within a camera’s field of view. This results in frequent false alerts caused by animals, weather conditions, falling leaves, passing vehicles, or reflections from windows. In fact, it’s estimated that in the U.S. for example, 90% of intruder alarms are false.

AI eliminates these inefficiencies, distinguishing between relevant and irrelevant factors based on predefined criteria.

What Events Can it Filter?

Some systems are more advanced than others and offer different functionality, but here are some examples:

Recognising suspicious objects and masked individuals.

Detecting the early signs of a fire faster than traditional smoke detectors.

It can be set up to ignore items below a certain size – for example, small animals.

Analysing behaviour to detect anomalies. For example, it can differentiate between a person approaching a restricted area and a cat wandering through the scene. It can differentiate between someone loitering or an innocent passer-by.

Improving access control. Thanks to facial recognition, it verifies whether unauthorised personnel are trying to access restricted areas.

It can be set up to only monitor certain zones within the camera’s field of vision – again, giving ARCs more control over the relevance of alerts.

We’ll look at some other specialised use-cases in a moment.

Core Technology Components

The intelligence behind such systems is thanks to several key technologies.

Computer Vision

This form of AI allows cameras to process visual information for analysis. Multiple moving objects can be tracked across complex scenes.

Video Analytics

Video analytics uncovers insights from raw video by analysing movement patterns, behaviour, and contextual information. Trajectory analysis involves mapping and interpreting movement paths, while anomaly detection identifies activities that deviate from established baseline patterns.

Multi-Spectral Analysis

For specialised use-cases, there’s multi-spectral processing – integrating data from infrared, thermal, and other visual sensors.

Standard surveillance systems work well under controlled conditions but factors like low lighting, shadows, smoke, dust or unstable backgrounds make it harder to get accurate insights. Another challenge is when important elements are a similar colour to the background. Including additional data streams helps solve these problems.

What is the AI Event Filtering Process?

Here’s the typical process by which the technology operates.

1. Data Collection and Processing

Video feeds are captured and processed. No special security cameras are needed to use AI event filtering – it’s typically compatible with standard hardware.

However, some companies use edge computing, where data is analysed at the point of capture (the hardware) before it’s sent to any central system. This reduces the load on the network, potentially reducing latency.

The use of IP-based security systems and cameras is on the rise, influenced by the expansion of IP infrastructure and the upcoming PSTN switch off. These provide an alternative to edge computing in terms of transmission speed and reliability.

2. Advanced Object Detection

The algorithm identifies objects then classifies them. Tracking multiple objects helps in interpreting complex scenes.

3. Contextual Analysis

The system analyses behavioural patterns against established baselines. Factors like business hours and scheduled activities provide further context.

4. Threat Evaluation and Response

Finally, a risk assessment determines potential threat levels. Based on pre-set parameters, the AI decides whether the event poses a genuine threat. For example, a person loitering near a restricted area at night may trigger an alert, whereas a tree swaying in the wind will not.

If an event is classified as a security threat, an alert is sent to the alarm receiving centre for human verification and response.

How Does AI Event Filtering Help Alarm Receiving Centres?

Reduction of False Alarms

As mentioned, this is one of the main benefits. Conventional motion detection systems can trigger hundreds of false alerts daily. AI event filtering minimises these occurrence which leads to faster response times and greater efficiency.

Greater Situational Awareness

The classification of objects, behaviour, and potential threat levels gives security personnel more information to use when making decisions – helping them make decisions faster.

Improved Resource Allocation

With fewer false alarms to manage, staff can focus on verifying and responding to real threats. Of course, this reduces operational costs and improves overall performance.

As reported by the FPA, UK government data from 2016/17 – 2021/22 showed that 97.5% of fire alerts were false. When it comes to fire detection specifically, video analytics offers a valuable supplementary approach to identifying potential incidents, even though it won’t eliminate false alarms from other detection methods. But thanks to visual confirmation and additional context, these systems can help staff more effectively distinguish and verify genuine fire threats, improving overall detection accuracy.

Integration with Existing Video Alarm Monitoring Systems

AI event filtering systems can integrate with existing security infrastructure, including cloud based alarm monitoring software. It improves the capabilities of surveillance cameras, alarm systems, and access control mechanisms without requiring an overhaul of the entire setup.

Other Use-Cases of Integrating AI Event Filtering with Alarm Receiving Software

Commercial and Retail Premises

In these settings, AI can differentiate between normal customer activity and suspicious behaviour such as hiding items within clothing. This can apply on the shop floor and at point-of-sale, when software solutions are integrated with POS cameras. After-hours monitoring is also enhanced, protecting premises from vandalism and break-ins.

Safety-Critical Operations

Facilities like airports and power plants need the most stringent security measures. Event filtering helps monitor restricted areas for trespassers, unauthorised vehicles, and other suspicious activity that presents a risk.

Public Safety

Public spaces can be monitored for suspicious activities or unusual crowd behaviour. This includes detecting items such as weapons and abandoned packages.

More advanced algorithms can assess factors like crowd density and behaviour patterns, predicting issues like overcrowding.

Cameras can track vehicles entering and existing premises and detect lingering or attempts to access restricted areas. Of course, this improvement to perimeter security applies in many private scenarios as well.

Beyond Security

Video analytics can help detect environmental risks – for example, flood detection is possible when cameras monitor rising water levels.

Gas leak detection is possible by combining thermographic cameras with advanced analytics. Changes in temperature and heat patterns can then be detected accurately, increasing the safety of industrial settings.

Then there are applications that improve efficiency in smart cities, such as traffic management. The options are comprehensive.

Summary

AI event filtering is the result of combining computer vision with video analytics. It reduces false alarms, improves situational awareness, and boosts the efficiency of alarm receiving centres.

Beyond monitoring for criminal activity, it can assist with detecting fires, gas leaks and other environmental threats. It also helps with traffic monitoring and other non-security applications.

No specialist hardware is required – you don’t need to replace your CCTV cameras to get the benefits. In many cases, AI event filtering solutions will integrate with your existing equipment and security alarm monitoring software.

One example is Calipsa, a leading AI alarm filtering provider. GeminiSense now integrates with Calipsa Detect, their object detection solution which reduces false alarms by 93 – 99%. To learn more, contact us today.