Image

For more than 40 years, AH-64 Apache helicopter crews have flown low-altitude attack and reconnaissance missions to support U.S. Army ground troops. The Apache is among the most sophisticated aircraft in the Army’s inventory, requiring pilots to manage complex sensor, navigation and defense systems, often while operating at night and in difficult terrain.

With small, fasting-moving drones emerging as a key threat to both helicopters and ground forces, researchers at the University of Arizona Wyant College of Optical Sciences are working to help Apache crews see a wider battlefield and distinguish details and threats more effectively.

Researchers in the Infrared Imaging Group, led by optics professor Ronald Driggers, have developed two prototype sensor systems designed to expand the sensors' field of view while improving their ability to detect drones. Doctoral student Chance Lawrence is contributing to the testing and evaluation of the systems. 

Each prototype contains 21 sensors mounted in a fixed ball, which is placed on the nose of the aircraft. One ball is equipped with uncooled long-wave infrared sensors and would be used on the Apache for flying at night; the other uses visible-spectrum cameras and serves as a testbed for research on imaging characteristics.

“Each of these cameras is stitched together to form one large contiguous field of view,” Lawerence said. “A partial field of view of the live image is then passed directly into the pilot's headset – a small window inside what is effectively a giant panorama.”

Image

While the conventional Apache pilotage system uses a single thermal camera that swivels on a gimbal to follow the pilot's line of sight, the new sensor is designed to add a wide-field layer to the helicopter's existing pilotage and targeting sensors – broadening situational awareness while reducing the physical limitations of current displays.

“Having the ball eliminates the moving gimbal and provides more coverage to both soldiers in the cockpit, allowing them to use it simultaneously,” Lawrence said.

These sensor systems are also designed to be more compact and less expensive, while still meeting performance demands. Additionally, the team replaced the standard monocle display with a binocular head-mounted display made by defense technology company Elbit Systems. Together, these optimizations provide a 210-degree horizontal field of view, more efficient passive threat detection, enhanced depth perception and greater sensor redundancy.

The approach could give Apache’s pilotage sensors a field of view five to seven times wider than current pilotage systems that can be used to search for drones and combat vehicles, Driggers said. Both prototypes are currently undergoing preliminary flight tests, which have been successful.

Visual and thermal prototypes

Funded through federal appropriations supported by the University of Arizona and members of Arizona's congressional delegation, the project is managed by the Apache program office, PM Apache, with the Army Research Laboratory providing research oversight. 

The effort builds on decades of infrared imaging and sensor research conducted at the University of Arizona.

Each prototype operates in a different portion of the electromagnetic spectrum. The visible-spectrum system uses complementary metal-oxide-semiconductor, or CMOS, cameras – the same type of sensors found in smartphones. These cameras detect reflected sunlight or moonlight and perform best during the day. In contrast, the thermal system relies on uncooled long-wave infrared sensors that detect radiation caused by heat emitted directly from objects.

Image

“Anything above absolute zero emits electromagnetic radiation,” Driggers said. “All objects and people emit infrared light that can be detected. With the thermal ball, pilots can fly in zero-visible-light conditions like it's daytime."

The uncooled design makes the new thermal system more practical for operational use. Traditional high-performance thermal cameras require cryogenic cooling systems to suppress radiation generated by the sensors themselves, Lawrence said. Those systems are expensive, heavy and complex to maintain.  

Designed by U.S. based defense contractor Leonardo DRS, the uncooled sensors that Drigger’s team implemented operate at room temperature, reducing cost and complexity while remaining suitable for high-performance tasks.

Refining sensor performance 

One challenge with uncooled sensors involves reducing image distortion known as fixed-pattern noise, or FPN. Unlike random static that changes from frame to frame, FPN remains fixed within the sensor’s pixels and can introduce misleading information that interferes with target detection. 

Lawerence’s research examined how FPN behaves when scenes are in motion and how it affects the ability to distinguish real terrain, obstacles and moving targets. 

Building on previous findings that motion can actually improve target visibility in noisy imagery, his work developed a framework for accounting for this effect when evaluating pilot performance in sensor models. Those insights helped to inform the development of the long-wave infrared pilotage system.

Lawrence also determined the longest integration times for the visual cameras that still allow for effective flight operations. Integration time is the maximum amount of time each pixel spends collecting light during a frame. Longer integration times can improve image quality by gathering more signal, but they can also introduce blur and lag. 

Working with Apache pilots, Lawrence validated the upper limits before imagery becomes too smeared or delayed for safe flight. Although this research tested the visible cameras, the results translate to the uncooled infrared sensors, which have similarly slow integration times.

Detecting drones with artificial intelligence

Beyond hardware optimizations, software must coordinate all of Apache capabilities and systems simultaneously and without delay.

The sensor and display systems must combine feeds from multiple cameras, correct image distortions, render imagery into the pilot’s headset and overlay flight information – all while reserving enough computing power to detect drones in real time. 

Image

To streamline this process, the team is exploring artificial intelligence and deep-learning algorithms that can automatically identify drones specifically for the new system. Training those algorithms to perform reliably remains an ongoing effort.

“Imagine there is a drone outside against the sky. That’s not too hard for the system to detect,” Driggers said. “Put the drone against clouds. That’s a little harder. But put it in front of a mountain, and it becomes much more difficult for the system to rapidly distinguish it."

While existing onboard instruments give the Apache the “senses” to detect guided munitions actively targeting it, the pilotage ball gives the helicopter “eyes” to see the world around it, Lawerence said. 

The ball acts as another layer of defense by reconfirming what the Apache’s warning systems flag. It can look for more subtle threats like small drones that emit none of the signals that older systems were built to detect. With integrated AI, the sensors on the ball could also begin flagging these threats automatically, extending what the existing system can detect or reaffirm.

“Altogether, this is going to be a powerful upgrade," Driggers said.