A meta-imager, also known as a front-end camera, developed at Vanderbilt University, shows potential to replace traditional scanning optics in machine-vision applications by capturing images more rapidly and efficiently.
By nanostructuring glass materials, the bulky optical camera is transformed into a meta-imager filter, enhancing front-end processing to encode information more efficiently. These imagers work in conjunction with advanced backend systems to offload computationally intensive tasks to high-speed, low-power optics. The resulting images have diverse applications in government, defense, healthcare, and security sectors.
In a Nature Nano publication, Professor Jason Valentine, from the electrical engineering department and deputy director of the Vanderbilt Institute of Nanoscale Science and Engineering, along with colleagues, present the proof-of-concept meta-imager.
The team includes Xiamen Zhang, a doctoral researcher in mechanical engineering, Hanyu Zheng, a Ph.D. candidate, and Yuankai Huo, an assistant professor of computer science. Quan Liu, currently a doctoral associate at MIT, and Ivan I. Kravchenko, a key R&D team member at the Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, also contributed.
The meta-imager’s architecture, as described by the authors, bridges the gap between physical and digital systems, offering potential applications in machine vision, data security, and artificial intelligence due to its compact design, high speed, and low energy consumption.
To facilitate data transmission for specific classification tasks, the team optimized an optic system comprising two metasurface lenses. They trained networks on datasets of handwritten numbers and commonly used images for testing various machine learning algorithms. The meta-imager achieved an accuracy of 98.6% for handwritten numbers and 88.8% for clothing images.
