Conquer False Alerts in a Security Monitoring System
It’s the age-old tale of the shepherd boy who cried “Wolf!” too many times – his sheep became wolf snacks. Desensitised by too many false alarms, the villagers ignored the boy’s pleas for help, even though the danger had actually materialised. Comparable to false alarms in a video monitoring system, the effectiveness of an operator and the response unit to act on the alarms can be severely compromised by aptly-named ‘nuisance alerts’.
Initially security surveillance was mainly live viewing of camera footage. However, the proliferation of surveillance cameras has created a deluge of footage that is impossible to monitor. Motion trigger and line crossing technology provides some filtering, however the false alarm rate even with these technologies can still be excessive. Too many false alarms can create contentious relationships between homeowners, businesses, central monitoring stations and first responders.
In an industrial context, under normal conditions a monitoring operator is exposed to an average of 3.1 alerts per hour, which can increase to as many as 52.8 alerts per hour. Just imagine the stats in a security monitoring scenario!
- This means that the operator has to perform almost an action per minute under unusual circumstances.
- Aside from alarm management tasks, the operator also has to communicate with the first response unit, and this mental workload becomes overwhelming quickly, compromising the ability to respond to genuine activations and impacting operator performance quality.
- This makes it critical to ensure that the information presented to the operator is manageable, otherwise there is an increased risk of error.
In video surveillance systems with very active scenes e.g: thick vegetation or extreme weather conditions, false alerts can become overwhelming. For example, in a solar park with 50 cameras running video analytics more than 150 alarms are detected in a single day. On rainy or windy days, these numbers can spike to 500 alarms daily, the vast majority of which are nothing to worry about. So which alerts should we be worried about in a security monitoring context? Without accurate information, it’s difficult to know so it’s necessary to treat them all as worrisome but this comes with some alarming consequences.
- Disruption to operating procedures: Every alarm, including false alarms needs to be evaluated and addressed which can result in an increased manpower demand in the control room. To meet such demand, resources must either be diverted from core tasks, or recruited externally to handle an increasing load of verification processes or veracity investigations. This is especially true in respect of response logistics. Responding to false alarms is an extremely wasteful operation.
- Greater likelihood of genuine alarm events going unnoticed: The risk of intruders or dangerous threats going unnoticed increases, and the potential for damage or threat to life is escalated.
- Irreparable reputational harm: The security response company runs a high risk of reputational damage due to failure to perform, which can negatively impact business viability. The more critical the outcome, the more potential damage the business stands to suffer in terms of a publicity backlash and financial damages.
- Confidence in security effectiveness is compromised: Where failures to detect occur, customer confidence in choice of security provider is eroded.
- Rising crime levels: Where operators have more information than they can handle, they’re likely to miss critical events. Just as worrisome – an operator may correctly identify an alarm, once it’s already too late. Alarm overload, inadequate response time and nuisance alarms all play a part in crime statistics.
Where perimeter protection with real-time alerting is the aim, the security industry is wary of the effectiveness of video analytics. Such markets expect analytics to deliver near 100% accuracy in order to consider them useful. Without this accuracy, monitoring operators grow weary of the ceaseless false alarms and will eventually disable the alerts.
Video analytics has come a long way since its first debut in 1998. Now, we’re looking at sophisticated cloud-based video analytics that makes use of software-based artificial intelligence (AI), and machine learning along with deep learning to filter out up to 95% of nuisance alarms. The repetitive, time-consuming task of identifying false alarms on feeds from thousands of cameras is now performed by AI software, leaving human operators free to focus on genuine alarms.
Depending on the camera setup, and the complexity of the surveillance scene, a single operator can handle up to 500 cameras. Once integrated into the security control room’s operating systems and customer cameras, AI-based false alarm reduction software has a number of immediate benefits:
- Reduce nuisance alarms by up to 95%
- Enhance monitoring operator performance
- Respond only to genuine threat events
- On-board more customers without increasing headcount
AI-powered video analytics work as an additional layer that delivers alerts into the security company’s existing Video Management System or Alarm Monitoring System. An accurate artificial intelligence object detection system is applied to dramatically reduce false alarms while viewing and actioning alerts in the user interface of the company’s choice. While most video analytics software like motion detection and object detection rely on rule-based algorithms to test numerous hypotheses before making a prediction outcome, sometimes there are simply too many variables in video footage to allow for high-quality nuisance reduction. That’s where neural networks come into play to vastly improve accuracy of false alarm reduction software, working to classify motion event images for future use. Images pooled from cameras in the field are sent up to the cloud for detailed analysis in the neural network. In the classifying model a deep neural network is trained on more than 100,000 images that have been banked over time. This training takes place with human oversight, in order to ensure that objects are correctly labelled. This deep neural network is essentially a black box of mathematical equations which are repeated in different layers, drawing features from the images and classifying them according recognisable objects. Put simply, this classification engine is a box that takes images on one side and provides a classification based on probabilities out the other side.
Internally, this deep neural network contains tens of millions of parameters which are tuned through a model training process to provide accurate classifications for the datasets provided. Depending on the architecture used and the computing used to run the image classification, several million several million images can be processed in a day which makes it possible for the false alarm reduction software to perform deep analysis on images from camera feeds with an increasingly accurate outcome. In this way, false alarm reduction software takes in the lessons picked up through the neural network, which results in an accuracy that only improves with time and human oversight.
- Fewer false alarms to attend to will have an overall cost reduction effect across the operational chain, in addition to a more efficient use of human resources.
- It’s not an investment in a static product, but rather an investment in a dynamic system that will keep learning and improving its performance the longer the software is in play.
- Acts as an additional layer of alarm verification that works in monitoring environments to expand human capabilities while working to eliminate human error at the same time.
- Security companies can be assured that they’re providing their customers with an increasingly accurate and effective service by augmenting human experience with technology.
As economic conditions get tougher and crime levels inevitably rise, consumers’ security budgets will be impacted – this means that their security providers need to earn their subscription premiums or justify their fee increases by showing real value and reliability. Security companies are going to find it exceptionally difficult to grow their market share without resorting to cost-cutting measures, and increasing efficiencies through false alarm reduction software is one of the most effective ways to cut costs without negative impact on customer service.
We’re not making outrageous claims when we say that DeepAlert can reduce false alarms in your security operations by 95% because we’re 100% capable of proving it.
To put us to the test, get in touch today to ask about your FREE demo.
Related Articles
No categories assigned to this post.
