UK combat engineering specialist Pearson Engineering, known universally for its mine-clearing hardware, is now adding a software-driven approach to address the mine threat. The company used the Eurosatory 2026 defence exhibition, held in Paris from 15 to 19 June, to showcase Threat-Sense: its new artificial intelligence (AI)-driven landmine detection system.
A fully passive system, Threat-Sense delivers enhanced battlefield awareness by enabling ground vehicles (crewed or uncrewed) or unmanned aerial vehicles equipped with a high-resolution RGB camera to scout ahead of ground forces, identifying and mapping explosive surface-laid threats in real time.
“This is a counter to the most present mine threat in Ukraine, which is hastily placed surface threats, either by personnel, missile distribution, or drones,” Pearson Engineering business executive Wilf Sergant told Warsight at the show on 16 June. “That’s the threat we’re countering because that’s the threat that is being seen in Ukraine in the modern battlespace.”
The system can use any high-resolution RGB camera system (1080p resolution or better), the job of which “is purely to feed back real-time RGB imagery,” explained Amish Patel, Pearson Engineering’s principal engineer for human/machine teaming. “You take that imagery in real time, feed it into Threat-Sense, and Threat-Sense analyses every single image that’s been fed in from that camera and, using the neural network that we’ve trained, it spots if there’s any objects of interest in that image. And it does that for every single image coming in, so if you had a camera running at 30 fps, which is fairly common on a drone, for example, 30 times a second, our AI-driven inference engine, our AI neural network, is scanning those images for threats.
“The way it does that is by using deep learning methods, which is a subset of AI, where we train the neural network using loads and loads of images containing the objects that are principles that we want to capture,” said Patel. “The way that Threat-Sense is designed is unique compared to other systems in that the data pipeline we’ve created, which generates imagery to train the AI, is 100% synthetically generated, so this avoids the need to capture loads and loads of images in real life, which would typically be the case when you go and train an AI-based object detection system.”
Patel noted that Pearson has built a fully synthetic training environment within which every aspect of an environment, such as the weather, terrain conditions, locations of objects of interest, can be controlled. “Within a few hours, we’re able to generate 50,000 images, which are highly photo realistic and introduce no bias to the AI — such as if we were only generating images of days where it’s sunny, for example – so we can validate the data that we’re generating before we train the AI, and that’s a really important step in improving performance.”
As a result, Threat-Sense can achieve a greater-than-80% positive detection rate and a less-than-10% false alarm rate.
“The product itself is designed for use in combat engineering operations, such as minefield breaching, but also on demining operations more on the humanitarian side,” said Patel. “How it would be employed would be for missions such as minefield reconnaissance purposes, using the AI within Threat-Sense as a tool to detect where surface-laid threats may exist, so that commanders can make better decisions for upcoming missions.”
He added that the addition of infrared cameras in addition to RGB ones is currently in the system’s development pipeline.
Sergant added that Threat-Sense’s deep learning technology makes it highly scalable. “In real terms, it means that we can scan in a threat and load it onto the system quickly, instead of gathering hundreds of thousands of images of that threat, so it’s very scalable in the capabilities we can offer to an end user. If they come back to us and say we’re experiencing an uptake or increasing usage in ‘X threat’, then as long as we have a surrogate that we can scan in, we can load that onto the system.”
Regarding how Threat-Sense moves forward commercially, Sergant told Warsight, “We’ve been quite successful in getting it to the hands of users for trials, so we’ve been on NATO trials, trials with the British Army; that has not been an issue because it’s quite an interesting capability for them. It is about tailoring our approach to meet their specific needs, and so we are very open in our relationship with them to build the best possible product or capability that we can.”
Patel added, “The relationship that we’ve had with the British Army during testing has been really good for refining the system. From an engineering perspective we can only design to our best intent, but ultimately we need to talk to the end user to understand how that capability is going to be deployed and advance it to get to the point where they’re comfortable in using it.
The company also confirmed it is open to joint ventures to move Threat-Sense forward.
“For decades, in missions such as minefield breaching within the combat engineering realm and demining within the post-conflict humanitarian side, the operators and the soldiers have been exposed to significant risk, given the explicit nature of the threats that we’re trying to deal with,” Patel noted. “So with the adoption of newer technologies, starting with robotics, how to remove the soldier or the operator from harm’s way, we have technology that allows for that. We’re then using AI as an additional layer of technology or intelligence to feed back information to commanders and operators so that they can better plan for mission success. And then we have autonomy, which is the last layer. So, the system of systems approach is where individual capabilities can work together in a scalable manner, so that operators and soldiers can shift from being the guys on the ground, doing what we describe as ‘dull, dirty, and dangerous’ work, are becoming decision makers, instructing autonomous assets to go away and do tasks that would expose them to significant harm otherwise.”
