Deep‐Learned Broadband Encoding Stochastic Filters for Computational Spectroscopic Instruments. Issue 3 (27th January 2021)

• Record Type: Journal Article
• Authors: Song, Hongya; Ma, Yaoguang; Han, Yubing; Shen, Weidong; Zhang, Wenyi; Li, Yanghui; Liu, Xu; Peng, Yifan; Hao, Xiang
• Is Part Of: Advanced theory and simulations. Volume 4:Issue 3(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-01-27
• Subjects: computational spectroscopy -- deep learning -- hyperspectral imaging -- optical inverse design; Science -- Simulation methods -- Periodicals; Science -- Methodology -- Periodicals; Engineering -- Simulation methods -- Periodicals; Engineering -- Methodology -- Periodicals; 507.21
• DOI: 10.1002/adts.202000299
• Languages: English
• ISSNs: 2513-0390

Abstract

Abstract: Computational spectroscopic instruments with broadband encoding stochastic (BEST) filters allow the reconstruction of the spectrum at high precision with only a few filters. However, conventional design manners of BEST filters are often heuristic and may fail to fully explore the encoding potential of BEST filters. The parameter constrained spectral encoder and decoder (PCSED)—a neural network‐based framework—is presented for the design of BEST filters in spectroscopic instruments. By incorporating the target spectral response definition and the optical design procedures comprehensively, PCSED links the mathematical optimum and practical limits confined by available fabrication techniques. Benefiting from this, a BEST‐filter‐based spectral camera presents a higher reconstruction accuracy with up to 30 times enhancement and a better tolerance to fabrication errors. The generalizability of PCSED is validated in designing metasurface‐ and interference‐thin‐film‐based BEST filters. Abstract : Computational spectroscopic instruments have great potential in integration, cost‐effectiveness, and portability. However, the optimization and the inverse design of suitable filters are still challenging. Here, a novel deep‐learning‐based design method is proposed. It offers an end‐to‐end optimization pipeline for both the filter and software design, and improves the design accuracy significantly.