Artificial intelligence drives analytics trends in 2020

Video analytics has been a top trend in video surveillance for years, and in 2020 it continues to be a hot topic. More and more end users are requesting video analytics as part of its video surveillance solution, and artificial intelligence (AI) is helping to improve false alarms rates and make analytics smarter. This does not come without challenges, but video analytics providers say they are up to the task.

AI analytics the new standard?

As AI and deep learning become more accepted, its application to video analytics are helping create more reliable, efficient and effective data. The Physical Security Business 2019 to 2024 report by Memoori projects that AI video analytics will become standard for video surveillance solutions in the next 10 years. More immediately, Memoori believes AI technology applied to video surveillance will become mainstream by this year.

Kevin Waterhouse, Managing Director of VCA Technology expects to see a rise in AI and deep learning applied to security, enabling video surveillance systems to ‘’learn’’ what a potential threat may look like (e.g., a vehicle or a person). “Deep learning and AI will empower businesses to drive efficiency among their security teams, ensuring resources aren’t wasted and improving efficacy,” he added.

John Kedzierski, SVP of Video Security Solutions at Motorola Solutions also pointed to the ability of AI to reduce the amount of work required to process the abundance of data available. Manually extracting data and formulating insights is a time-consuming process prone to human error. Utilizing AI analytics can reduce hours of work to minutes, saving both time and resources

AI and deep learning can also help analytics improve both accuracy and detection rates, helping it tackle one of its biggest challenges: false alarms. This could significantly reduce false alarms and create more secure environments, while also improving the efficiency and productivity of both video surveillance installs and monitoring elements, explained Waterhouse. That is why he believes there will be more businesses adopting video analytics enhanced with AI and deep learning within their current systems in 2020.

The move of analytics to the edge is also expected to grow in importance within surveillance applications, giving users access to analyzed data in real time. Andreas Conrad, Head of Marketing at Qognify noted that it makes sense to run many types of analytics on the edge, as it reduces bandwidth consumption — there are also instances where a combination of edge and server-based functionality works very well. A good example of a combination solutions is when the camera is able to crop a face from the video stream and automatically sends it to the server for more advanced analytics to be applied.

Improving ROI

The increasing adoption of video analytics is expected to drive more conversations about generating a bigger return on investment (ROI). Traditional surveillance systems are rarely monitored in real-time. Waterhouse explained that choosing a modern system with analytics capabilities drives operational efficiencies by creating real-time alerts to which security staff can promptly respond, thereby driving better business outcomes.

Analytics enhanced by deep learning and AI will also help more companies utilize surveillance cameras for more than just security — it will allow them to obtain applicable business intelligent insights. “For example, video analytics enable cameras to be deployed across diverse scenarios and will provide businesses with a greater return on investment thanks to multiple applications such as, investigating customer behavior, counting footfall and better optimizing space and resources,” Waterhouse added.

Concerns about face recognition

Face recognition is everywhere nowadays, yet despite its acceptance in commercial offerings like smartphones, on other fronts it still remains a major topic of debate. Around the world regulations and concerns regarding face recognition vary drastically. Itsik Kattan, CEO of Agent Video Intelligence (Agent Vi) believes the controversy around face recognition will peak during 2020. 

“After concerns around privacy increased during 2019, and some countries/cities banning or restricting the use of face recognition technologies, this will become a topic that the industry will have to formally tackle in 2020,” Kattan said. “Vendors and integrators will have to choose whether to adopt such technologies, consider to what extent they need to be a party to regulating the use of such technologies, and no less important, to understand how the commoditization of face recognition technology affects the business sustainability of companies dealing purely with face recognition development.”

Adoption of face recognition in video surveillance solutions, however, is growing, despite concerns about its invasiveness and fears of inappropriate use. This may lead it to be used only in specific cases under strict guidelines, which could complicate the "business flow" for face recognition vendors.

Protecting privacy

Privacy is another major concern for end users, especially with AI analytics and face recognition being rapidly deployed. When used correctly they are powerful tools that can increase the effectiveness of physical security systems and improve public safety. In the wrong hands, though, it could be a dangerous weapon. To gain public trust, analytics providers are dealing with these concerns in a variety of ways. 

Understanding that customers using AI-powered solutions can have a significant impact on society, Kedzierski pointed to how Motorola Solutions takes a more rigorous approach to the application of AI — beyond the fundamental tenets of fairness, privacy, understandability and reliability. 

“Motorola Solutions has a commitment to the responsible use of AI as well as individual privacy rights. Data stewardship is integral to these new capabilities, and we build compliance controls into our products to support this. Specific measures for many of our solutions include user authentication, password strength enforcement and lock-outs on multiple invalid attempts, audit logs of user actions, video redaction, automatic firmware updates, data retention periods and the ability to expunge or remove data on demand,” Kedzierski explained.

VCA Technology address the privacy issue by not storing any video footage in the first place, thus eliminating the need to protect stored data. Waterhouse explained that their solutions are engineered to analyze images, produce and forward metadata using secure channels, and subsequently discard the footage.

The key will be for analytics providers to work with lawmaking bodies to properly figure out proper frameworks and best practices for the use of analytics and protection of privacy. Analytics providers must also continue to develop and improve algorithms to reduce inconstancies and increase data protection, while also continuing to educate the industry and general public.