How AI Video Analytics Helps Reduce False Alarms

Monitoring centers receive signals and alarms from various sources, including video surveillance systems, motion detectors, fire and burglary alarms, and access control systems. However, not all alarms are genuine emergencies. Unwanted signals disrupt the normal functioning of monitoring systems, causing not only operational but also financial burdens for monitoring centers and end-users alike. In this article, we will delve into the impact of noise interference on the effectiveness of security systems and unveil why leveraging advanced AI technology in the form of video analytics makes an effective solution to decrease false alarms.

What is noise?

Noise in the context of monitoring centres is an unwanted and unnecessary alert that is sent to a security operator for assessment. When a camera feed is shown on the operator’s screen, the operator must view and acknowledge it. When the camera feed is not actionable and shows something that is not interesting, that is noise.

When monitoring a large camera network, which is often geographically distributed, there may be thousands of false alerts per day because of noise. Noise can overwhelm operators, resulting in missed genuine detections and significant costs.

Where does noise come from?

If a camera is showing a static scene, there is no value in monitoring it. Security operators only monitor cameras when there is a change. Motion activated cameras work only sending a feed to be reviewed if more than a specified percentage of pixels in the scene change.

Ideally, this change will identify a person or vehicle entering the camera view, which can then be assessed by a security operator. However, in most cases the pixels change from other causes, resulting in noise for the monitoring centre.

There are several potential triggers for motion-activated cameras that feed noise to monitoring centres.

● Animals (spiders, insects, rodents, pets and other animals)

From tiny spiders to pesky insects, curious rodents to beloved pets, buildings are home to a diverse array of creatures, which can sometimes cause unexpected motion triggers. Spiders can build webs across cameras and trigger motion every time they move. If the building has a rodent problem, these can also trigger motion, especially at night. Pets or wild animals such as racoons or squirrels are other culprits.

● Wind

Wind can move objects in front of the field of view. These could be flying objects such as pieces of litter or foliage, if there are trees in front of the camera.

● Lighting changes

Changes to lighting can result in significant number of pixels changing in a camera view, which is interpreted as motion. This can be caused by a flickering fluorescent light, or lights being turned on or off. If the camera picks up car headlights or reflections from the sun shining off cars, that also triggers motion.

If the camera is positioned to face outdoors, cloud cover can also cause lighting changes.

● Camera issues

Some cameras can be affected by interference from high voltage electrical cables or radio frequency, or incorrectly shielded cables. This may result in a grainy appearance.

Image noise can also be caused by poor illumination where the light level in the image is too close to the level of noise in the sensor.

What is the impact of noise?

Noise in a camera network causes severe financial and operational impacts on security monitoring center operations.

● Psychological impact on operators

Constant alerts from noise can result in psychological harm to operators from hypervigilance, resulting in stress, fatigue and damage to relationships.

● Staff turnover

If staff are stressed due to a barrage of alerts, they are more likely to leave. Staff turnover is financially costly due to recruitment and retraining costs. Turnover also disrupts security operations as new team members need time to fit in and “learn the ropes”.

● Missed valid alerts

With a very large number of false alarms from noise, operators may be so busy assessing those that they may miss some valid alerts. This may result in very significant customer complaints, given that they expect valid alerts to be detected and acted upon promptly.

● Loss of productivity

Every false alarm needs to be assessed and acknowledged, taking around 15 seconds per incident. This means that more operators are needed, just to review the false alarms and to identify the few valid ones.

Moreover, false alarms undermine public trust in monitoring systems, potentially leading to delays in real emergency response. This highlights the pressing need for an effective solution to reduce noise and minimize false alarms.

Reducing noise

Noise can be reduced (but not eliminated) by fixing the root causes:

● Spiders and insects – apply environmentally friendly low-grade insecticide to cameras. Regularly clean them with a soft brush to remove webs.

● Rodents – get a pest controller to address rodent issues using baits or traps.

● Pets – position camera views to avoid the areas frequented by pets.

● Wind on foliage – mask areas of foliage in software (where these are not in an area where people may be present).

● Wind-blown litter – clean up litter and secure bins.

● Lighting issues – repair or replace flickering light fixtures.

● Reflections – reposition the camera views.

● Camera electrical interference – shield high voltage cables. Use high grade coaxial cables.

● Camera sensor noise – provide adequate background lighting.

There are many sources of noise that can’t be easily fixed, especially when the cameras are on customer premises.

AI video analytics: a weapon against noise

Noise in monitoring centers is a pervasive issue that hampers the effectiveness of security systems. However, AI-powered video analytics software offers a robust solution to reduce noise. By leveraging powerful machine learning algorithms, AI software can intelligently analyze video feeds, filtering out false alarms caused by noise and accurately identifying legitimate threats. Thus, noise is minimized without removing critical feeds of interest.

Scylla’s False Alarm Filtering works by automatically identifying people or vehicles in camera feeds, impressively reducing the number of false positives by up to 99.95%. As a result, the time spent on noise is reduced, increasing productivity of monitoring operators, and reducing stress that can lead to psychological harm and employee turnover.

By integrating AI-powered video analytics software with existing monitoring systems, noise reduction becomes seamless, ensuring a smooth transition and enhanced accuracy.

Scylla’s ROI

Scylla has developed an online calculator to allow you to estimate the benefits from using Scylla False Alarm Filtering. This captures benefits from the productivity improvement from operators not having to spend time constantly assessing false alarms but doesn’t quantify some of the other benefits such as reduced employee turnover.