Rapid in-situ calibration of computational micro-spectrometer with few-shot meta-learning

• DOI: 10.1364/oe.522256
• Publication Date: 2024-04-29T16:00:27Z
• Authors: Meichen Yang; Xiuteng Chen; Xiaowei Wu; Qingfeng Hu; Yinpeng Chen; Zongyin Yang; Yusong Sheng; Yang Chen; Likun Han; Jiaqi Zhu; Mingzhong Pan; Shijie Liu; Hongxing Qi; He Zhu; Ning Dai
• Keywords: Monochromator; Imaging spectrometer

Computational micro-spectrometers comprised of detector arrays and encoding structure arrays, such as on-chip Fabry-Perot (FP) cavity filters, have great potential in many in-situ applications owing to their compact size snapshot imaging ability. Given manufacturing deviation environmental influence are inevitable, easy effective calibration for spectrometer is necessary, especially applications. Currently strategies based on iterative algorithms or neural networks require accurate measurements pixel-level (spectral) functions through monochromator large amounts standard samples. These procedures time-consuming expensive, thereby impeding Meta-learning with few-shot learning ability can address this challenge by incorporating the prior knowledge simulated dataset. In work, we propose a meta-learning algorithm free measuring function samples calibrate micro-spectrometer effectively. Our comprises 16 types FP filters covering wavelength range 550-720 nm. The center each filter type deviates from design up 6 After 15 different color data, average reconstruction error test dataset decreased 7.2 × 10-