The biggest trend and topic seen in the security industry this year is definitely artificial intelligence (AI) and deep learning. While the technologies aren’t particularly new, this year they have more than ever captured the attention of the market due to various factors.
The biggest trend and topic seen in the security industry this year is definitely artificial intelligence (AI) and deep learning. While the technologies aren’t particularly new, this year they have more than ever captured the attention of the market due to various factors: an increase in data that’s available for meaningful analysis, the emergence of hardware devices with high computing power, as well as the maturity of networking infrastructure for both landline and wireless transmissions. As such, this year's Security 50 companies are embracing the growth opportunity in AI.
“We are trying to develop AI technology-based products which can provide benefits beyond simple video analytics functions to users. We target to building a close-to-perfect surveillance management system. As we grow, our well-classified product range will be broadened, and it will become all answers to customers’ needs,” said Jimmy Park, Senior Director of Strategic Product Management Team at
Hanwha Techwin. “We are currently working on optimizing Al technology for camera, recorder and VMS. We set scenarios in order to implement optimized functions for each vertical market as well as to reflect user-requested functions. When high-performance AI technology is applied, various functions can be practiced and utilized throughout all vertical areas with less expected errors and fail results.”
AI-powered analytics
“Deep learning-powered analytics were a major trend which received a lot of interest and marketing activity this year. Increasingly vendors are launching smart network cameras and recorders with embedded deep learning capabilities,” said Josh Woodhouse, Senior Analyst for Video Surveillance at
IHS Markit. “Initial developments in this area in the second half of 2016 were fueled by semiconductor developments from Nvidia and Intel Movidius — now we see a much larger range of semiconductor vendors marketing deep learning for video surveillance. With the increased competition we expect prices will decline more rapidly, meaning the technology can begin to filter down from high-end cameras and recorders into lower-priced equipment.”
“The next step in video analytics is to dive deeper to gain very specific insights into video content, including analyzing human behavior through the use of neural network video analysis. Video will not only be used to track the usual movement of cars and people or detect items left behind, but will also be relied on more frequently to bring behaviors of interest to the attention of security personnel,” said Jammy DeSousa, Senior Product Manager for Security Products for Building Technologies and Solutions at
Johnson Controls.
“Magal recognizes the importance of AI and machine learning to the future of security. We are currently developing target classification capabilities for both our video management and perimeter intrusion detection solutions. We are also actively researching and developing other ways to add AI/machine learning functionality to our products,” said Brian Rich, CTO and Deputy CEO at
Magal Security Systems.
Facial recognition applications
“Our facial recognition solution uses AI and deep learning. The ability to harness this technology has led to increasing performance levels and accuracy of detection for this solution and we are looking at ways to include this in our other technologies,” said Marie Clutterbuck, CMO of
Digital Barriers.
One company that has achieved a lot of success with facial recognition is China-based Videopark Technology, which has a strong focus on the banking industry. “Our facial recognition has passed tests by public safety and other agencies. We are one of the industry pioneers to roll out the facial recognition-supported NVR, as well as the ATM smart alert DVR. Our ATM smart alert system currently sees the most applications in China,” said Luo Jun, Director and VP of
Videopark Technology.
“Milestone sees the artificial intelligence and machine learning technologies as creating hyper growth markets by enabling IoT software companies to aggregate large numbers of connected devices and integrate them to form a new smart systems automation market. Milestone is prioritizing our technology roadmaps to capture these hyper growth opportunities,” said Kenneth Hune Petersen, Chief Sales and Marketing Officer at
Milestone Systems, adding that several innovative technology investment areas will be delivered over the coming years.
“Within the connected devices market we have already made investments in enabling the Milestone platform to aggregate the largest numbers of connected devices. We now support over 6,000 device types, have migrated the video processing from CPU to GPU and have developed a metadata engine to capture the new information created from smart devices,” he said. “Within the smart systems market we are investing in technologies which leverage deep learning and automate processes and operations. We also see opportunities to use AI to enhance many operations. And we recently just announced that we intend to support the new Nividia Metropolis Platform to building AI smart cities within our soon-to-be-released new Video Processing Server.”
“We see a lot of development happening in AI and deep learning. We believe AI and deep learning will have a great role to play in the future. The overall solutions will be much more comprehensive,” said Yogesh Dutta, COO of
CP PLUS. “One area we’re focusing on is traffic where we’re are giving those kinds of solutions supported with license plate recognition. We are also using it in the enterprise segment, giving enterprise-level solutions for visitor management. They are also using facial recognition from a deep learning perspective, receiving an alert when somebody is coming.”
Avigilon, which has put a strong emphasis on AI, explains how it can help solve some of the challenges and difficulties facing users. “Through the power of AI, we’ve developed technology that better focuses human attention on what matters most — enabling users to answer the critical who, what, where and when of an investigation with decisive action. For instance, innovative new technology like Avigilon Appearance Search technology, a sophisticated deep learning AI search engine that sorts through hours of footage with ease, allows users to quickly locate a specific person or vehicle of interest across all cameras on an entire site,” said Willem Ryan, VP of Global Marketing and Communications at
Avigilon. “Avigilon Appearance Search technology is designed to make searching for a person as easy as searching the internet. It incorporates the unique characteristics of a person’s face to quickly locate a specific person of interest across an entire site. Using deep neural networks for face signatures, this technology is designed to increase the speed and accuracy of investigations.”
Broader perspective
Axis Communications, meanwhile, has a broader perspective on this. “Machine or deep-learning is mostly used for video analytics, but I expect the technology will be an important component in many different applications and products in the future. Over time it will become a common tool for software engineers and will be included in many different environments and devices,” said Johan Paulsson, CTO of
Axis Communications. “However, the surveillance industry has a history of sometimes overpromising with video analytics, and we are especially conscious of that when it comes to deep learning. We think deep learning has to mature further before it is ready for market in a broader perspective.”