Industry analysts agree artificial intelligence will have a huge impact on physical security market growth, particularly video surveillance.
Erin Harrington, contributing writer for SSI magazine, interviewed several top executives to weigh in on the subject:
Artificial intelligence (AI) and deep learning technologies are coming out of the shadows of engineering labs and military applications and finding their way into consumer and commercial security deployments. With the ability to tap into other information facilitated by sensors, such as people counting, identity management and heat monitoring, an AI system can provide end users infinitely more information than traditional security applications ever could.
But despite the high hopes and sometimes dashed expectations, AI systems are still very much a foreign term to the security market. A more educated and realistic understanding of AI capabilities will only be made possible by the benefits of time and experience.
One thing it doesn’t take advanced analytics to figure out: AI technologies will only continue to evolve and take a predominant place in the video surveillance market.
A recent report from Research and Markets notes that AI will be a key driver of growth in the global physical security market. It projects a CAGR of 7.3% from 2018 to 2023. In addition, AI video analytics, according to Memoori, are expected to propel growth in the video surveillance market and see a CAGR of 13.4% between now and 2023.
[Here are excerpts from thought leader] insights on the AI landscape [that] speak to the latest developments, security and safety applications, early adopter markets, longer term projections, systems integrator opportunities, and overall challenges.
Advanced Video Analytics Highlight Latest Developments
Jake Cmarada, senior business development, enterprise sales for video surveillance camera maker Dahua, says that continued advancements in facial recognition, perimeter detection and heat map analysis are among the latest AI developments. The ability to house enormous image libraries and process from edge devices is now the trend. Deep learning models on the server side allow data collection and access in real-time to make safer, smarter and more efficient sites.
“Realistically, in security we can now only see examples of machine learning,” states Tim Palmquist, vice president Americas at VMS provider Milestone Systems. “Trained algorithms combined with new compute capabilities are making the long-awaited promise of reliable and productive video analytics come to life and become practical in day-to-day use. We don’t yet see commercial security examples of software that can learn and evolve being utilized much. We should expect, however, that these types of applications will soon make their way into our marketplace and, in turn, unlock a whole new chapter of innovation and opportunity.”
End Users Reap Greater Business Intelligence
Some potential security customers are already experiencing how AI can impact their bottom line in other ways. Learning customer demographics for retail vertical applications, for example, to analyze the purchase data, gender, age and interests to provide related merchandise/product promotion is among one of the most common use cases Cmarada points to.
“AI-enabled software delivers insights and intelligence that streamline business processes and provide business intelligence to improve the whole operation and grow the business,” he says.
“There are many cases where complex algorithms can automate otherwise manual processes, but historically there was not enough compute capability to effectively deliver results,” Palmquist contends. “Recent advancement in GPU and CPU capacity have begun to resolve the compute issue, making machine learning more practical in day-to-day use cases. Today we see video analytic solutions that work very well. Color, direction, correlation, object identification, facial recognition, synopsis, etc. are all examples that we see successfully in use today.”
Object recognition, pattern recognition, anomaly detection, predictive analysis are some of the more applicable AI use cases Jack Wu, co-founder and CEO of drone system integration specialist Nightingale Security. He predicts these to become very popular soon: “Currently, we’re seeing object detection as a pervasive trend. Anomaly detection will increase in accuracy as object detection becomes more mature and less false positives are being recorded. In the near future, rule-based task automation will become the standard and lead to autonomous patrolling, autonomous incident reaction and on-board decision making. Those features will lead to robots being able to monitor and react to incidents in real-time with humans brought into the loop as the robot’s performing identification, tracking and following.”
Large industrial customers and critical infrastructures as well as the military rank among the earliest AI adopters, Wu believes, due to mission-critical facilities protection that, if left unsecure, can have large financial, political and national security consequences. Retail, utilities and government agencies are also among the early adopter crowd to target, according to Cmarada.
Taking it from another perspective, Palmquist notes, “In our industry, video analytics is the obvious early adopter of machine learning. Consequently, they will also likely be the first forward with learning software that better meets the definition of artificial intelligence. Outside our industry, we can see that autonomous driving capabilities — machine learning augmented by LIDAR [light detection and ranging] — is another early mover.”