Channeled imaging spectropolarimeter reconstruction by neural networks

Learnable sparse dictionary compressed sensing for channeled spectropolarimeter

Channeled spectropolarimetry enables real-time measurement of the polarimetric spectral information of the target. A crucial aspect of this technology is the accurate reconstruction of Stokes parameters spectra from the modulated spectra obtained through snapshot measurements. In this paper, a learnable sparse dictionary compressed sensing method is proposed for channeled spectropolarimeter (CSP) spectral reconstruction. Grounded in the compressive sensing framework, this method defines a variable sparse dictionary. It can learn prior knowledge from the measured modulated spectra, continuously optimizing its own structure and parameters iteratively by removing redundant basis functions and refining the matched basis functions. The learned sparse dictionary, post-training, can provide a more accurate sparse representation of the Stokes parameters spectra, enabling the proposed method to achieve more precise reconstruction results. To assess the efficacy of the proposed method, simulations and experiments were conducted, both of which consistently demonstrated the superior performance of the proposed approach. The suggested method is well-positioned to enhance the efficiency and accuracy of polarimetric spectral information retrieval in CSP applications.

Physics-guided neural network for channeled spectropolarimeter spectral reconstruction

A reconstruction method incorporates the complete physical model into a traditional deep neural network (DNN) is proposed for channeled spectropolarimeter (CSP). Unlike traditional DNN-based methods that need to employ training datasets, the method starts from randomly initialized parameters which are constrained by the CSP physical model. It iterates through the gradient descent algorithm to obtain the estimation of the DNN parameters and then to obtain the mapping relationship. As a result, it eliminates the need for thousands of sets of ground truth data, while also leveraging the physical model to achieve high-precision reconstruction. As seen, the physical model participates in the optimization process of DNN parameters, thus achieving physical guidance for the DNN output results. Based on the characteristic of the network, we designate this method as the physics-guided neural network (PGNN). Both simulations and experiments demonstrate the superior performance of the proposed method. Our approach will further promote the practical application of CSP in a wider range of fields.

Full-Stokes polarization multispectral images of various stereoscopic objects

Polarization multispectral imaging (PMI) has been applied widely with the ability of characterizing physicochemical properties of objects. However, traditional PMI relies on scanning each domain, which is time-consuming and occupies vast storage resources. Therefore, it is imperative to develop advanced PMI methods to facilitate real-time and cost-effective applications. In addition, PMI development is inseparable from preliminary simulations based on full-Stokes polarization multispectral images (FSPMI). Whereas, FSPMI measurements are always necessary due to the lack of relevant databases, which is extremely complex and severely limits PMI development. In this paper, we therefore publicize abundant FSPMI with 512 × 512 spatial pixels measured by an established system for 67 stereoscopic objects. In the system, a quarter-wave plate and a linear polarizer are rotated to modulate polarization information, while bandpass filters are switched to modulate spectral information. The required FSPMI are finally calculated from designed 5 polarization modulation and 18 spectral modulation. The publicly available FSPMI database may have the potential to greatly promote PMI development and application.

Polarized angle-resolved spectral reflectometry for real-time ultra-thin film measurement

We propose a polarized, angle-resolved spectral (PARS) reflectometry for simultaneous thickness and refractive-index measurement of ultra-thin films in real time. This technology acquires a two-dimensional, angle-resolved spectrum through a dual-angle analyzer in a single shot by radially filtering the back-focal-plane image of a high-NA objective for dispersion analysis. Thus, film parameters, including thickness and refractive indices, are precisely fitted from the hyper-spectrum in angular and wavelength domains. Through a high-accuracy spectral calibration, a primary PARS system was built. Its accuracy was carefully verified by testing a set of SiO₂ thin films of thicknesses within two µm grown on monocrystalline-Si substrates against a commercial spectroscopic ellipsometer. Results show that the single-shot PARS reflectometry results in a root-mean-square absolute accuracy error of ∼1 nm in film thickness measurement without knowing its refractive indices.

Convolutional neural network-based spectrum reconstruction solver for channeled spectropolarimeter

Channeled spectropolarimetry is a snapshot technique for measuring the spectra of Stokes parameters of light by demodulating the measured spectrum. As an indispensable part of the channeled spectropolarimeter, the spectrometer module is far from being perfect to reflect the real modulation spectrum, which further reduces the polarimetric reconstruction accuracy of the channeled spectropolarimeter. Since the modulation spectrum is composed of many continuous narrow-band spectra with high frequency, it is a challenging work to reconstruct it effectively by existing methods. To alleviate this issue, a convolutional neural network (CNN)-based spectral reconstruction solver is proposed for channeled spectropolarimeter. The key idea of the proposed method is to first preprocess the measured spectra using existing traditional methods, so that the preprocessed spectra contain more spectral features of the real spectra, and then these spectral features are employed to train a CNN to learn a map from the preprocessed spectra to the real spectra, so as to further improve the reconstruction quality of the preprocessed spectra. A series of simulation experiments and real experiments were carried out to verify the effect of the proposed method. In simulation experiments, we investigated the spectral reconstruction accuracy and robustness of the proposed method on three synthetic datasets and evaluate the effect of the proposed method on the demodulation results obtained by the Fourier reconstruction method. In real experiments, system matrices are constructed by using measured spectra and reconstructed spectra respectively, and the spectra of Stokes parameters of incident light are estimated by the linear operator method. Several other advanced demodulation methods are also used to demodulate the measured spectrum in both simulation and real experiments. The results show that compared with other methods, the accuracy of the demodulation results can be much more improved by employing the CNN-based solver to reconstruct the measured spectrum.

Reconstruction and calibration methods for a Mueller channeled spectropolarimeter

Channeled spectropolarimeter (CSP) measures spectrally resolved Stokes vector of light and Mueller matrix of sample from a snapshot. While reconstruction and calibration methods for Stokes CSP have been well established, their Mueller CSP counterparts are lacking. In this paper, we propose methods for Mueller spectrum reconstruction and Mueller CSP calibration. Mueller CSP is modeled as a modulation matrix, linking the Mueller spectrum to be measured and the modulated spectrum from the spectrometer. We describe an optimization problem to solve the Mueller spectrum, where both the regularizer and the residual threshold constrain the result, making our reconstruction accurate, efficient, and noise-robust. The Stokes spectrum generated by polarization state generator and the analyzing vector of polarization state analyzer are measured in situ, the convolution of which construct the calibrated modulation matrix of Mueller CSP. Total polarimetric errors and spectroscopic errors are treated as a whole and represented by the calibrated modulation matrix. Both imaging and non-imaging Mueller CSP are experimentally calibrated. Reconstruction results show high accuracy with a root-mean-square error (RMSE) of 0.0371. The proposed methods help make Mueller CSP practical and have the potential to be general reconstruction and calibration methods for imaging and non-imaging Stokes-Mueller CSP.