Real-Time Video Analytics: The 'Killer App' for Edge Computing
By Michael Martinez
 

people looking at many video screens

Big brother is going to need a lot of cloud and edge computing.

As surveillance cameras are projected to grow 20% worldwide every year for the next five years, they will need strict real-time analytics to mine video content, and that scenario will demand a system of public clouds, private clusters, and edges, according to a new study by analysts with Microsoft Research, which appears in the October 2017 issue of Computer.

There is a camera instailled for every 29 people on the planet, and in developed nations, the number rises to a camera for every eight people.

That means a lot of computing power will be needed to analyze traffic, security, crime, and consumers’ digital assistants.

The answer lies in “a geographically distributed architecture of public clouds, private clusters, and edges that extend down to the cameras is the only approach that can meet the strict real-time requirements of large-scale video analytics, which must address latency, bandwidth, and provisioning challenges,” the researchers say.

Geo-distributed video analytics infrastructure
Geo-distributed video analytics infrastructure. Each organization deploys their own private cluster of different sizes, and relies on the public cloud for additional capacity. Network links between cameras, private clusters, and the public cloud have diverse bandwidths (represented by the width of the arrow).

Read “Real-Time Video Analytics: The Killer App for Edge Computing.” (login may be required for full text)

“Because of high data volume, compute demands, and latency requirements, cameras are the most challenging ‘things’ in the Internet of Things. Thus, large-scale video analytics could well be edge computing’s ‘killer app.’ Tapping into the potential of recent dramatic increases in the capabilities of computer vision algorithms presents an exciting systems challenge,” the researchers say.

Their proposed video analytics system would demand little resources but offer high accuracy. In fact, one such system is now in 24/7 operation in Bellevue, Washington, where live cameras process feeds from traffic intersections. Traffic accidents are among the top 10 causes of fatalities.

“The system generates directional traffic volumes and raises alerts on anomalous traffic patterns. We are on track to identify dangerous conflict patterns to minimize traffic deaths. We also plan to expand our solution to cities across the US and worldwide,” the researchers say.

The emerging field of real-time video analytics will drive a wide range of applications.

Vision Zero for traffic

Bellevue Washington traffic cameras analytics
Production deployment details of traffic analytics in Bellevue, Washington, based on live traffic intersection camera streams. The figure shows a dashboard of multiple camera feeds being analyzed simultaneously, the video analytics pipeline, and charts of directional volumes.

Many cities have developed “Vision Zero” initiatives to reduce traffic deaths. That initiative developed in Sweden and seeks to eliminate all road fatalities.

“Detecting “close calls” between cars, bikers, and pedestrians helps city planners to identify hazardous areas in which preemptive safety measures are needed,” the researchers said.

Self-driving and smart cars

Sensors and cameras in high-end cars already keep motorists safe and help them make good decisions on braking, turning, and other maneuvers that may imperil them or other drivers.

“Output from traffic camera analytics can also trigger decisions in cars,” the study says. “A traffic light–mounted camera can detect hard-to-see pedestrians (for example, those standing between parked cars) and can warn approaching self-driving cars using dedicated short-range communication (DSRC).”

Surveillance and security

Government and corporate security constitute a large measure of video and camera surveillance. The United States, United Kingdom and China each use millions of cameras to monitor our activities on public sidewalks and roads.

Increasingly, even police officers are wearing body cams.

This area will continue to ripe for growth opportunities.

“Analysts predict that drone cameras will also significantly aid in home-surveillance activities,” the study says.

The Microsoft researchers who authored the new study are Ganesh Ananthanarayanan, Paramvir Bahl, Peter BodíkKrishna Chintalapudi, Matthai PhiliposeLenin Ravindranath, and Sudipta Sinha.

 


Michael Martinez

 

About Michael Martinez

Michael Martinez, the editor of the Computer Society’s Computer.Org website and its social media, has covered technology as well as global events while on the staff at CNN, Tribune Co. (based at the Los Angeles Times), and the Washington Post. He welcomes email feedback, and you can also follow him on LinkedIn.