Besides video analysis, what are the benefits of AI and machine learning?

Time:2020-11-23

Introduction:Artificial intelligence (AI) and machine learning (ML) have caused a sensation in the physical security market, which has promoted video analysis to a new level of accuracy. In fact, these terms have become the common buzzwords of the whole industry. But the potential for AI and machine learning to impact the physical security industry goes far beyond their ability to improve video analytics.

Besides video analysis, what are the benefits of AI and machine learning?

Artificial intelligence (AI) and machine learning (ML) have caused a sensation in the physical security market, which has promoted video analysis to a new level of accuracy. In fact, these terms have become the common buzzwords of the whole industry. But the potential for AI and machine learning to impact the physical security industry goes far beyond their ability to improve video analytics.

The theme of our expert group Roundtable this week is: how can artificial intelligence or machine learning benefit the physical security market in addition to better video analysis?

Nigel Waterton-Chief Revenue Officer, Arcules

While I do think it’s a long way from AI driven results in today’s movies, adding these algorithms can greatly help business leaders ultimately make better decisions and reduce risk. In addition to video analysis, this goal is also the core of the development of this segment. Adding another tool, such as cloud based capabilities, to this intelligence can bring additional advantages and additional flexibility that our industry has never seen before.

At the end of the day, the real benefit of this technology for the physical security industry is the ability to access the data coming in from a variety of Internet of things (IOT) devices and use that information to establish best practices for business operations, thereby strengthening their own strength and enabling organizations to better understand the risks faced by their organizations.

Per Björkdahl-Chairman, ONVIF

Users can leverage video analytics with greater efficiency and accuracy by using AI, or more specifically, through deep learning and machine learning capabilities. Although these terms are sometimes used interchangeably, each has a different advantage. Machine learning can provide better and more accurate event detection and analysis. When people think of video analysis, they usually associate it with face recognition. However, machine learning is much more than that. It can monitor movements and processes, and detect traffic and events.

In contrast, AI is used to mimic what a person can do and help improve some low-level tasks. AI in physical security is designed to complement the end of human capabilities. AI helps improve automatic decision-making and alerting.

Sean Foley-SVP, National Accounts, Interface Security Systems LLC

We are excited about the AI revolution in video analytics, but AI is not limited to video. The real power of AI is to process large, often different sets of data to generate viable insights. For example, asset protection professionals have a deep understanding of what point of sale transactions are dangerous signs of fraud. AI can take this understanding to an exponential level, evaluating millions of transactions among thousands of employees to identify fraud early in the process, even before it occurs, thereby reducing shrinkage.

At the same time, the same type of AI pattern recognition can be applied to reduce the false alarm of central station, or to make ultra accurate prediction of system failure, so as to improve customer service. Our industry is just beginning to pair almost incomprehensible data with AI engines and algorithms. The application is unlimited and customers will benefit from it.

Stuart Rawling-Vice President of Technology and Customer Engagement, Pelco, Inc

The real possibility of improving intelligence through deep learning and other AI driven technologies applied to video is that in the long run, we don’t start watching videos until after things happen. The goal of this high level of intelligence through video collection may be achieved automatically, so that the security operator does not need to make the decision needed to respond. Instead, the intelligence driven next step will be automatically communicated to all stakeholders – from the on-site security to the local police / fire department.

Instead, when security executives access the video corresponding to the event, it’s because they want to see the event themselves. Isn’t automation, the ability to simplify response, and instant response the goal of a holistic, data rich monitoring strategy? Yes for almost all businesses.

Aaron Saks-Product and Technical Manager, Hanwha Techwin America

In addition to better video analysis, artificial intelligence (AI) or machine learning can greatly benefit the physical security market. For cameras, AI can not only eliminate false positives through motion based analysis, but also do more. From automating tasks to running routines and comparing data, artificial intelligence and deep learning have the potential to change the way we use security cameras.

With more cameras installed than humans can monitor, to take advantage of all this information, we need AI to understand the new data we’re collecting and tell us what to pay attention to. We want to know what’s unusual: is that car going the wrong way on the street? Is there anyone in the middle of the road? These devices are powerful new IOT sensors that directly enhance business and operations.

Adam Wynne-Software Engineering Manager, Security and Safety Things GmbH

Artificial intelligence (AI) and machine learning can also benefit the physical security market through improved access control systems and integration of resulting data with other devices. By using this technology, the algorithm can identify individuals through biometrics and automatically integrate them with security cameras to develop more comprehensive access control solutions. AI can enhance the biometric fingerprint system by improving the recognition speed and accuracy. In addition, artificial intelligence and machine learning also bring the added benefit of real-time detection of complex events, which used to be part of forensic analysis only after fact analysis. This makes the physical security system and response more simplified and complex.

Jonathan Moore-Product Director, AMAG Technology, Inc.

Video analysis is often used to identify people and other objects and then trigger specific actions, such as opening doors or triggering alarms. Although this function is useful, data analysis has great value and can provide useful insights from the large amount of data stored in the access control system. Artificial intelligence can “learn” the typical access patterns of each user and warn security when suspicious or abnormal behaviors that may pose a threat to the organization are detected.

In addition to detecting potential hazardous activities, data analysis can also be used to better understand the occupancy and flow patterns of buildings, to help implement physical alienation, to highlight misconfigured or potentially faulty panels and devices, etc. Data analysis programs can help enterprises improve their security and internal threat procedures, understand their facility usage and traffic patterns, and optimize their security hardware.

John Davies-Managing Director, TDSi

Artificial intelligence (whether real artificial intelligence or complex machine learning) has great potential in assisting physical security. By learning and improving its own data, AI can quickly determine what is normal or abnormal behavior, so as to detect potential problems early. The benefits of video analysis have been well documented, but AI in centralized security systems can monitor a wider range of complex data. For example, in a busy airport or railway station, a central AI system can handle the movement of people in and out of safe areas (using access control and video surveillance) and find patterns that may indicate congestion problems or suspicious behavior.

In addition, we also see more and more artificial intelligence being used in cutting-edge technologies such as UAVs, which can determine whether there are any problems or needs attention in remote installation, power line or gas pipeline without manual guidance.

Brian Baker-Vice President, Americas, Calipsa

Artificial intelligence and machine learning have brought exponential changes to the way physical security processes inputs from cameras and sensors. The data is the fuel for AI, and the camera provides a lot of video for viewing. AI’s deep learning algorithm can automatically detect differences between human and vehicle movements, rather than animals, blowing leaves or reflecting light. The result is a significant reduction in false positives and potential associated fines.

We see AI as an added layer of security, helping, not replacing, humans to better protect people and assets. With artificial intelligence, operators at a central monitoring station or enterprise security operations center can focus on real alarms to improve security response. By reducing the time wasted by false alarms, managers can extend operations without increasing staff. Today, cloud based AI software solutions add their capabilities to compatible cameras almost anywhere in the world.

summary

Artificial intelligence (AI) and machine learning provide useful tools for understanding a large amount of Internet of things (IOT) data. By helping to automate low-level decisions, these technologies can make security operators more efficient. Intelligent features can extend integration options, such as access control to increase the use of biometrics. AI can also help improve monitoring mechanisms and processes. Intelligent systems can help end users understand the occupancy and traffic patterns of buildings, and even help achieve physical distances. These are just a few possible uses for technology – in the end, everything is possible.

Link to the original text: https://ai.51cto.com/art/202008/623391.htm
The article is transferred from 51CTO. All the viewpoints in this paper have nothing to do with the circle of machine intelligence technology