The following piece is from insight.tech, an Intel-sponsored publication leveraging the Intel® Partner Alliance to provide business and technical decision-makers the latest and greatest technology trends and business solutions in the IoT space.
Fixed-to-mobile, AI-powered video analytics for buildings
By Pedro Pereira. November 16, 2022.
Smart buildings deliver continuous streams of data from sensors, cameras, and other devices to optimize operations and respond to user needs. With AI-powered video analytics, building operators can easily process this data. But these building solutions are usually within permanent structures supporting fixed surveillance systems. Many locations, including construction sites, temporary work sites, parks, event venues, and other in-field settings, need mobile surveillance units because they are typically not equipped with video cameras or analytics.
One company facing this situation turned to Megh Computing, a video analytics solution provider, about a Mobile Surveillance Unit (MSU). It needed an advanced, cost-optimized, AI-based video analytics solution to monitor 5,000 mobile sites, all on solar power.
Existing MSUs typically use expensive “smart cameras” with analytics that detect intrusion. Since the AI models used in these devices are not very robust, they generate many false positives, which increase operational costs in monitoring the alarms and the bandwidth required for the backhaul.
Megh Computing introduced a new MSU, with an Intel® Core™ i3 NUC processing streams from four standard cameras on the unit. The system uses advanced deep learning models based on an Intel® Distribution of OpenVINO™ Toolkit-supported inference engine for detection of people and objects. Prabhat K. Gupta, CEO of Megh Computing, says, “Since these models deliver high operational reliability, Megh is able to virtually eliminate incidents of false positives, reducing operational costs for monitoring.”
The MSU’s video analytics pipeline is highly optimized and controlled by a motion filter that reduces overall power consumption for the system to less than the required 15W. This system also generates savings from lower communication and camera costs, resulting in compelling ROI.
“You want to be able to monitor any situation or crowd to see what’s happening. With an MSU, you can quickly roll the thing up, put it up, and start monitoring in conjunction with security staff,” Gupta says.
Customized Smart Building Solutions
Megh Computing leveraged its expertise with its VAS (Video Analytics Solution) for Smart Buildings to create the mobile option for its customers. VAS is a highly customizable solution that can target different hardware architectures and can be deployed from the edge to the cloud. VAS processes data from both cameras and sensors. It has been deployed in a variety of environments, including smart cities, smart warehouses, and retail locations.
VAS applies AI deep learning to accurately detect people and objects, and also anomalies in human behavior, to reduce physical risks. This was an important aspect to bring to its MSU solution.
“Obviously, there’s an explosion of data at the edge, and most of the data is video,” says Gupta. “And, except for data used to respond to security incidents, most of it is never analyzed.”
Companies that invest in costly video systems want to be able to analyze streaming data and gain business insights from it. For instance, how many people use a company’s spaces at any given time, and for what purpose? The answers can lead to operational improvements, such as better traffic flow and smarter security controls. Think of a retail kiosk. By monitoring the space, Megh’s solution can determine how long people wait in line, so the retailer can add another kiosk to avoid lost sales.
“We can also monitor space usage,” says Gupta. “We can see how people are spending time in work areas throughout the day. That helps with facility planning, for example.”
Reduce False Alarms with AI-Based Security Solutions
Besides offering customization and mobility, Megh Computing addresses a common problem with video analytics: the high rate of false alarms. A security firm turned to Megh when it was experiencing this problem.
After piloting Megh’s solutions side by side with their existing platform, the company made the switch. “The operator said we were basically able to eliminate all of their false positives,” Gupta explains.
Megh Computing minimizes false alarms through the use of its advanced AI deep learning models and its ability to continuously update these models for improved accuracy, using its continuous training framework. “This is a unique capability that assures that when the system issues an alert, the person who receives it can feel confident it’s real and take appropriate action to prevent a security incident,” says Duncan Moss, Principal Engineer at Megh Computing.
For instance, a car dealership uses the platform for after-hours monitoring. If someone acts suspiciously on the lot, intelligent cameras can guess the person’s intent by tracking their movements and alerting a security guard. An individual who moves from car to car, stopping for a few seconds by each vehicle, may plan to break in and steal a car. “That’s an example of how we are using a technology for behavioral analysis and prevention,” Moss says.
End-to-End Advanced Security Solutions
Megh Computing not only looks at situational monitoring of spaces but aims to address cybersecurity in its solutions. “People who are trying to break in can break in physically or electronically. If you want to provide complete security for an enterprise, you’ve got to look at both aspects,” Megh CEO Gupta says.
For example, VAS addresses the cyber aspect by monitoring network packets for congestion and signs of distributed denial of service (DDoS) activity. Thus, Megh provides a single platform for both physical security and cyber security threats.
To deliver all these capabilities, Megh Computing partners with Intel for its advanced AI hardware and software platforms. Megh’s solutions are certified as Intel Market Ready.
Going forward, Megh plans to continue leveraging the Intel relationship, not just for technology but also to widen its market reach. Having gained traction in smart buildings, Gupta says the company now targets retail and smart warehouses, too.
As it does, Megh will continue to look for more opportunities to deploy MSUs. After all, security and safety shouldn’t be confined to buildings; it’s needed wherever there is risk.