The paper "Deep-learning-based on-chip rapid spectral imaging with high spatial resolution" by PhD student Jiawei Yang was published online in Chip on April 7, 2023 (DOI: 10.1016/j.chip.2023.100045).

Spectral imaging extends the concept of traditional color cameras to capture images across multiple spectral channels, and is widely used in many fields such as remote sensing, precision agriculture, biomedicine, environmental monitoring and astronomy. Conventional spectral cameras based on scanning methods suffer from low acquisition speed, large volume and high cost. On-chip spectral imaging based on broadband modulation of metasurfaces and computational spectral reconstruction provides a promising scheme for the realization of consumer portable spectral cameras. However, the existing point-by-point iterative spectral reconstruction algorithm has the problems of long computing time and mosaic phenomenon in reconstructed spectral images.

In this work, we propose to combine deep unfolded neural network ADMM-net with freeform shaped metasurface spectral imaging chip to achieve on chip rapid spectral imaging with high spatial resolution. The network is used for spectral image reconstruction of real scenes. The reconstruction of a single 256×256×26 data cube takes only 18 milliseconds, and the reconstruction speed is about 5 orders of magnitude higher than the traditional point-by-point iterative spectral reconstruction algorithm. The mosaic phenomenon of spectral image is effectively eliminated. Furthermore, video-level dynamic spectral imaging was realized for the driving scene, with an imaging rate of 36 frames per second, which is expected to solve the problem of metamerism recognition in the automatic driving scene, and show the application potential in machine vision, industrial process control and other fields.

Jiawei Yang, Kaiyu Cui,* Yidong Huang,* Wei Zhang, Xue Feng, and Fang Liu, Deep-learning-based on-chip rapid spectral imaging with high spatial resolution. Chip 2023, 100045.

https://doi.org/10.1016/j.chip.2023.100045