Case studies

Create Safer Industrial and Utilities Plants with Intelligent Edge-Based Vision Solutions

Factory scene with blue-outlined smoke, juxtaposed with text promoting the 'Intelligent Edge AI System PE1100N Powered by NVIDIA® Jetson Orin.

Background: The need for a smart vision-based system to detect leaks, flares, fires, and other hazards

Safety is the primary factor in all industrial and utility plants. It is an overarching set of guiding processes and principles that provides a common framework protecting people, preventing catastrophe, and meeting compliance across operations.

However, industrial and utility plants are complex environments where many activities carry inherent risks that demand constant vigilance and careful management. These facilities handle hazardous materials, operate under extreme conditions, and contain equipment that can fail in ways that threaten worker safety and regulatory standards. These risks include the storage of flammable liquids in large onsite vessels, pressurized systems that can develop leaks or sprays, gas handling operations where invisible releases may occur, flare systems requiring emission monitoring, and fire-risk environments where early smoke detection is critical.

This is where US-based computer vision specialists VisionAery come in. Its founders recognized that many industrial and utility facilities were full of video feeds monitoring such threats, yet almost none of that footage was being used for real-time automation or safety decisions. The company had a mission: to close the gap between the collection of footage from operational scenarios and the ability for companies to respond quickly and intelligently to any emerging threats.

“We knew there was a huge market opportunity for intelligent vision solutions for operational problems through the development of an AI model that could turn ordinary cameras into smart sensors that detect leaks, flares, fires, and countless other hazards with high accuracy before they become costly events,” says Kaylor Greenstreet, Chief Executive Officer at VisionAery.

The image displays an interface showing the detection of a flare, which is outlined by blue lines. On the right side of the interface, there are detection instructions and data related to the flare.

Flare Monitoring

The image shows a spray coming from a machine, outlined in green on a detection interface. Instructions and data are displayed on the right side of the frame.

Spray Detection

“For example, liquid leaks are a significant concern within industrial settings and can potentially lead to dangerous and costly situations. Through training of over 400+ hours of 100% real leak videos, our system has been developed to be the world's most accurate and effective liquid leak solution, reducing risk across industrial sites. Similarly with smoke detection, our solution continuously scans every video frame in real-time, capturing the faintest wisp of smoke - even in the hardest‑to‑monitor areas - so companies can act before a spark becomes a shutdown.”

Water leakage outlined in green near pipes and tanks; instruction panel on the right shows 'Puddle Percentage: 10%, Spray Percentage: 0%'.

Leak Detection

Waste facility with orange lines marking smoke detection and a small orange circle highlighting fire; instructions and detection statistics displayed on the right.

Fire & Smoke Detection

However, many industrial and utility plants are located in remote areas, often with poor connection to cloud-based services. This meant the need for edge-native vision solutions engineered specifically for remote, bandwidth-constrained, and security-sensitive environments. “Our approach was to create a solution that let cameras stream locally to edge servers, with analytics running on-site with no dependence on cloud bandwidth or internet connectivity, keeping critical infrastructure air-gapped and data sovereign,” says Kaylor Greenstreet.

Solution: Deploying the intelligent edge AI system to improve industrial safety

The VisionAery solution leverages real-time computer vision on edge-connected cameras for the detection of liquid leak, liquid spray, fire and smoke, gas, and visible vapour. It can also be used for flare monitoring, helping oil and gas operators comply with regulations by monitoring black smoke emissions, delivering actionable notifications when flaring exceeds acceptable thresholds.

The system uses existing cameras connected to ruggedized edge computers that run AI computer vision models. The cameras capture real-time footage, which is processed locally by the edge computers using specialized AI algorithms trained to recognize specific hazards. When a threat is detected, the system immediately triggers automated alerts and integrations with existing plant systems - all without sending data off-site.

VisionAery’s computer-vision experts can partner with end users to design, train, and deploy custom AI models tailored to unique operational or safety challenges. It supplies the proven workflows, edge-ready frameworks, and hands-on guidance to steer the requirements and own the IP, shortening time-to-value without the steep learning curve that holds back many AI deployments in industrial environments.

“What we developed was a 100% edge-based solution, with camera-agnostic input, that could be easily integrated into existing systems,” says Kaylor Greenstreet. “This was exactly the approach that the market needed – giving end users real-time oversight of their operations with complete ownership of the data coming from our applications.”

At the heart of the system are rugged edge computers that run real-time AI-inferencing. VisionAery selected the ASUS IoT PE1100N Intelligent Edge AI System, powered by NVIDIA Jetson Orin™ NX and NVIDIA Orin™ Nano, which provides the high AI processing power required, with up to 100 trillion operations per second (TOPS) - delivering up to 5 times the performance of previous-generation devices. The PE1100N features a fanless design and diverse I/O, ready work across Wi-Fi, Bluetooth, and 4G/5 G via optional modules, providing the flexible connectivity required.

The PE1100N also features a compact design, measuring 152 x 114 x 72 mm and weighing 1.4 kg, and can operate in a wide range of temperatures and humidities. It is designed to meet military-grade MIL-STD 810H compliance for vibration and shock, making the computers suitable for installation in harsh environments, and ensuring the computers can withstand transportation to remote locations over long distances by truck.

“It was this combination of extremely high processing and inferencing power and ruggedisation, in a compact footprint, that made the PE1100N the perfect solution for our vision-system needs,” says Kaylor Greenstreet. “The PE1100N Intelligent Edge AI System provides fantastic performance on all levels, and in all conditions out in the field. We have deployed many modules in a wide range of industrial and utility plants, indoors and outdoors, and we haven’t had one fail yet.”

The outcome: VisionAery and ASUS IoT working together over the long term

The PE1100N comes from the ASUS IoT industrial product portfolio. Technical teams from VisionAery and ASUS IoT worked closely together to ensure the edge computers matched or exceeded all required specifications. Additionally, as a sub-brand of ASUS and a trusted NVIDIA partner, ASUS IoT is backed by a global presence and excellent inventory management, giving VisionAery confidence that supply would remain reliable even as demand increased.

“The ASUS IoT support team has been excellent throughout the development of the PE1100N, overcoming any deployment challenges and making sure the computers were well-suited for our specific needs,” says Kaylor Greenstreet. “For example, our deployments in remote locations can mean hardware is sometimes fed by fluctuating power from a renewables-based supply. The ASUS IoT team made sure the PE1100N could operate in all parameters, ensuring reliability in any environment.”

Michael Hou, product manager for Industrial motherboards, NVIDIA Jetson™ Series edge AI computers, and edge servers at ASUS IoT, says the partnership has been highly successful from the outset. “We ensured we had suitable technical resources in place to meet VisionAery’s specific needs. It is about having the right product, in sufficient volumes, to help the VisionAery solution scale in global markets.”

Both companies have an eye on the future. VisionAery is looking to leverage ASUS IoT's extensive portfolio of NVIDIA Jetson modules, ranging from the Nano to the higher-specification NX and AGX, and is excited about ASUS's collaboration with NVIDIA on next-generation chips. Higher-performing AI edge computers will enable VisionAery to run heavier vision language models on edge devices, raising the system's accuracy to over 99%. This will allow customers to do even more with their vision systems, thereby further increasing safety and efficiency.

“Hazard detection applications pose many challenges for hardware selection, such as harsh environments and requirements for stable supply, quick and scalable deployment, expertise consultation, and long-term cooperation,” says Kaylor Greenstreet. “We believe ASUS IoT, both as a brand and its hardware, can perfectly satisfy these needs.”

Michael Hou at ASUS IoT concludes: “We are excited to see how VisionAery will continue to advance AI edge-based vision systems and look forward to partnering - now and in the future.”