With funding from the U.S. Army, a Mizzou Engineering team and collaborators are using artificial intelligence in hopes of revolutionizing infrared sensors.
The work builds upon research previously conducted by Mahmoud Almasri, associate professor of electrical engineering and computer science, and Edward Kinzel, an associate professor of aerospace and mechanical engineering at the University of Notre Dame. The duo was awarded a patent for their metasurface integrated microbolometers — which incorporate networks of nanoscale antenna elements to provide spectral and polarization selectivity to the infrared sensor — in 2022.
With the new grant, the team will use machine learning to enable the metasurface integrated microbolometers to extract chemical environment imaged by thermal cameras. The infrared is sometimes called the molecular fingerprint region because chemical information is encoded in the spectra absorbed or emitted by many common substances. While hyperspectral infrared cameras exist, they are large, cooled and expensive.
“The goal is to create a system that is analogous to the way humans and other creatures perceive color,” Kinzel said. “This will provide similar competitive advantages for differentiating materials in the environment.”
“The objective of the project is to create a multispectral infrared detector that uses AI and suitable training, to recognize specific materials,” Almasri said. “We want to enable a user to make rapid decisions about their environment.”
In addition to military applications, the proposed system could be used in manufacturing or agriculture. The sensors could also improve how unmanned aerial vehicles and robots detect and respond to their surroundings.
This grant allows Almasri and Kinzel to bring in expertise from Associate Professor Derek Anderson who will use artificial intelligence (AI) to help them optimize their device.
“Dr. Almasri and Dr. Kinzel are designing next generation sensors. But it’s a gigantic design space with complex metrics. Different metasurfaces form different images, which is what a human or AI algorithm ‘sees’,” Anderson said.
In other words, the materials the team uses to design the sensors will affect and enhance the ability of AI learn the response of specific materials and identify them.
“Objects/materials emit different infrared radiation,” Anderson said. “Our eyes and brains have evolved to recognize materials based on three simple color channels; red, green, and blue. But, what’s the best set of infrared channels to robustly detect materials? We are exploring a framework that trains an AI based on a provided design from Dr. Almasri and Dr. Kinzel. The AI then explores and learns across different designs to help make new recommendations. Ultimately, the AI is designing the metasurface and sensor, not the human. The sensor is tailored to the application at hand.”
Ultimately, the team hopes to develop a device that’s affordable, lightweight and small enough to be integrated into a phone camera assembly.
Help revolutionize technology. Apply to the University of Missouri College of Engineering today!
The post Team using artificial intelligence to revolutionize infrared sensors appeared first on Mizzou Engineering.