AI AT THE EDGE

Every manufacturer wants to create a safer working environment for its employees. But achieving this can be difficult—and costly. Reducing risk in a factory means constant monitoring to identify environmental safety issues and proper workplace precautions. But this kind of vigilance can be labor-intensive.

“On average, a 10,000-square-meter factory has to have a minimum of 10 safety personnel, with two HSE (Health, Safety, and Environment) officers required for video supervision and monitoring,” says Stephen Li, CEO at Aotu, an AI company providing machine vision-based health and safety solutions.

That represents a significant investment for manufacturers, especially as industrial facilities need round-the-clock supervision. In addition, manual monitoring has its limitations, “Plant safety personnel work hard—but they’re stretched thin. They can’t detect most health and safety problems immediately, and they’re never going to cover 100% of the scenarios,” says Li. “Plus, there are delays in responding to issues in a timely fashion, since a human being has to actually make a phone call or go to the site of a safety violation in person to observe or correct it.”

It’s a challenging situation for manufacturers, who are committed to worker safety but are also under pressure to tighten budgets and optimize processes. But new machine vision health and safety solutions may provide an answer that keeps workers safe and satisfies the demand for greater efficiency.

Machine Vision Automates Safety Monitoring in Bottling Plant

Deployment at a bottling plant in China based on Aotu’s machine vision solution is a case in point.

The plant is operated by a major beverage company. The sheer size of the facility means many different areas to monitor, including rooftops, ceilings, boilers, waste areas, warehousing facilities, and more. Plus, the bustling site is filled with workers performing a wide variety of tasks, making supervision of employee behavior a complex undertaking.

In collaboration with Intel, Aotu developed a machine vision-based health and safety solution designed to analyze video feeds from the bottling plant and automatically alert safety personnel when an issue is detected.

The system’s AI algorithms are configured to monitor for environmental safety issues. The deployments cover nearly 1,000 key supervision points within the factories. At the same time, AI also analyzes video feeds for behavior-based safety violations: failure to wear proper protective gear, unsafe climbing and walking, unauthorized access to high-risk areas, violations of maximum occupancy limits, and so on.

If the system detects a problem, it captures a 30-second recording of the safety issue, classifies it as either a major or minor emergency, and sends an alert to a human supervisor for verification and response. If the problem is serious enough, a safety official can remotely trigger an on-site alarm and warning message to alert workers to imminent danger. For less severe issues, safety personnel have the option to follow up later for resolution and worker training.

After implementing the solution, the bottling plant saw an increase in both the efficiency and the efficacy of its safety program. “The use of AI reduced the workload of the plant HSE staff, and it ensured that safety issues were no longer going to be ignored,” says Li. “In addition, safety awareness among front-line workers improved significantly.”

#MachineVision health and safety solutions are gaining traction among large manufacturers—and their adaptability, cost-effectiveness, and ease of deployment should make them attractive to #SystemsIntegrators and smaller industrial businesses as well. Aotu via @insightdottech

Flexible Platform for Video Analytics

For a machine vision solution to be broadly useful to the manufacturing sector, it must be adaptable. A bottling plant, after all, is quite different from an auto parts factory, a high-tech fabrication site, or a chemical facility.

To create a robust yet flexible machine vision platform for industrial health and safety, Aotu decided to partner with Intel. Together, the companies were able to leverage the capabilities of a number of Intel hardware and software tools:

• 11th Generation Intel^® Core^™ processors offer optimization and acceleration for deep learning, AI, and machine vision scenarios.
• Intel^® Iris^® X^e GPUs are particularly well-suited to computer vision tasks such as smart video processing.
• Intel^® Xeon^® scalable processors enable configurations that require heavier workloads, and are also suitable for use in harsher industrial settings due to their ruggedized design and wide operating temperature range.
• The Intel^® OpenVINO^™ Toolkit provides pre-trained AI inferencing models and reference models for common industrial safety scenarios—as well as a foundation for the rapid development of custom AI algorithms.

“Intel’s processors and AI toolkits are a proven, powerful platform for building industrial AI applications,” says Li, adding that Aotu’s collaboration with Intel “helped us shorten our development time and design customizable, reliable low-code and no-code AI solutions for our end users.”

Towards Safer and More Efficient Industry

Machine vision health and safety solutions are gaining traction among large manufacturers—and their adaptability, cost-effectiveness, and ease of deployment should make them attractive to systems integrators and smaller industrial businesses as well.

But beyond the health and safety benefits, the inherent flexibility of these solutions will open other use cases as well.

As Li explains: “Fully integrated and optimized for advanced AI computing with the chipset partners, our machine vision solution has many potential smart manufacturing applications, including process digitalization, logistics automation, and predictive analysis, and will synergize well with emerging technologies like edge computing and 5G.”

In the future, look for computer vision to further the digital transformation of manufacturing in new and innovative ways, making Industry 4.0 safer, more efficient, and more profitable for all.

This article was edited by Georganne Benesch, Associate Editorial Director for insight.tech.