Monocular complex amplitude imaging via a polarization-multiplexed liquid-crystal-lens-informed Fourier neural network

Indoor airflow field reconstruction using physics-informed neural network

Modeling of the hysteretic behavior of nonlinear particle damping by Fourier neural network with transfer learning

Predicting positon solutions of a family of nonlinear Schrödinger equations through deep learning algorithm

Simulating convective heat transfer using traditional numerical methods requires explicit definition of thermal boundary conditions on all boundaries of the domain, which is almost impossible to fulfill in real applications. Here, we address this ill-posed problem using machine learning techniques by assuming that we have some extra measurements of the temperature at a few locations in the domain, not necessarily located on the boundaries with the unknown thermal boundary condition. In particular, we employ physics-informed neural networks (PINNs) to represent the velocity and temperature fields while simultaneously enforce the Navier-Stokes and energy equations at random points in the domain. In PINNs, all differential operators are computed using automatic differentiation, hence avoiding discretization in either space or time. The loss function is composed of multiple terms, including the mismatch in the velocity and temperature data, the boundary and initial conditions, as well as the residuals of the Navier-Stokes and energy equations. Here, we develop a data-driven strategy based on PINNs to infer the temperature field in the prototypical problem of convective heat transfer in flow past a cylinder. We assume that we have just a couple of temperature measurements on the cylinder surface and a couple more temperature measurements in the wake region, but the thermal boundary condition on the cylinder surface is totally unknown. Upon training the PINN, we can discover the unknown boundary condition while simultaneously infer the temperature field everywhere in the domain with less than 5% error in the Nusselt number prediction. In order to assess the performance of PINN, we carried out a high fidelity simulation of the same heat transfer problem (with known thermal boundary conditions) by using the high-order spectral/hp-element method (SEM), and quantitatively evaluated the accuracy of PINN’s prediction with respect to SEM. We also propose a method to adaptively select the location of sensors in order to minimize the number of required temperature measurements while increasing the accuracy of the inference in heat transfer.

Surrogate models leveraging Neural Networks (NNs) have demonstrated significant potential in accurately modeling the intricate dynamics of the injection molding process, enabling the prediction and optimization of process and material parameters on internal thermodynamics and final product quality. However, the substantial effort required for data collection remains a notable challenge. This study investigates the effectiveness of pretraining methods with varying fidelities for a two-dimensional cooling problem in injection molding. Specifically, NNs are pretrained using low-fidelity (LF) solutions derived from the Finite Element Method (FEM) and Physics-Informed Neural Networks (PINNs). These pretrained models are subsequently fine-tuned with high-fidelity (HF) simulation data to enhance prediction accuracy across previously untrained scenarios. While both FEM and PINNs provide approximations of the cooling process, PINNs offer a continuous solution space, however, at a higher computational cost compared to FEM. Respectively, FEM is more computationally efficient but less proficient at handling highly parameterized input spaces. Comparing different fidelities reveals that prioritizing the capture of main effects over detailed physical implementation during pretraining can be advantageous. The fine-tuned surrogate model demonstrates robust approximation capabilities with limited simulation runs, highlighting its data efficiency.

We propose the use of the binary amplitude of a diffracted field to obtain complex-amplitude encoding of a phase-only hologram. Although the complex amplitude in the hologram plane consists of an amplitude and phase, a phase-only hologram retains only the phase information of the complex amplitude, which degrades the quality of the holographic reconstruction. The proposed method approximates the amplitude in the hologram plane by a binary amplitude, which allows us to record the complex amplitude of the original optical field in a phase-only hologram. We verified the effectiveness of the method by applying it to two examples: a hologram reconstruction and the generation of Hermite–Gaussian beams.

Division of focal plane (DoFP) polarization image sensors capture polarization properties of light at every imaging frame. However, these imaging sensors capture only partial polarization information, resulting in reduced spatial resolution output and a varying instantaneous field of overview (IFoV). Interpolation methods are used to reduce the drawbacks and recover the missing polarization information. In this paper, we propose residual interpolation as an alternative to normal interpolation for division of focal plane polarization image sensors, where the residual is the difference between an observed and a tentatively estimated pixel value. Our results validate that our proposed algorithm using residual interpolation can give state-of-the-art performance over several previously published interpolation methods, namely bilinear, bicubic, spline and gradient-based interpolation. Visual image evaluation as well as mean square error analysis is applied to test images. For an outdoor polarized image of a car, residual interpolation has less mean square error and better visual evaluation results.

Modeling coupled groundwater flow and reactive transport for multi‐query tasks is computationally prohibitive, and standard Physics‐Informed Neural Networks (PINNs) require costly retraining for each new parameter. We introduce the Multi‐Physics Generative Pre‐trained PINN (MP‐GPT‐PINN), a meta‐learning framework to resolve this bottleneck. Its innovation is a two‐stage strategy: an offline pre‐training stage uses a greedy algorithm to build a compact library of PINN solutions for key parameters. A rapid online stage then generates new solutions by combining these pre‐trained bases in an optimal linear manner. For multi‐physics problems, we designed a parallel dual‐network architecture to couple the seepage and chemical fields robustly. Validated across complex groundwater systems, MP‐GPT‐PINN accelerates online predictions by four orders of magnitude while maintaining high fidelity, with dimensionless relative errors for physical fields approaching 10 −3 . This efficiency breakthrough makes computationally intensive tasks, such as large‐scale parameter inversion and uncertainty quantification, feasible in geosciences.

Research on optical encryption system based on unequal modulus decomposition and polarized vortex optical holography

Light fields contain rich information, including intensity, spectrum, and polarization, but conventional imaging systems are often monofunctional for spectral or polarization measurements, limiting their application scenarios. This work proposes a spectro-polarimetric imaging system that integrates computational spectral and polarization imaging to simultaneously capture both types of information. Using a division of focal plane (DoFP) polarization camera and 16 polarization-independent broadband filters to modulate and capture spectral-polarimetric content, the system reconstructs data at four polarization angles (0°, 45°, 90°, 135°) across 31 spectral bands (400-700 nm, 10 nm intervals) with a high spatial resolution of 2448 × 2048, achieving the full spatial-resolution of the polarization camera and enabling the extraction of the three Stokes parameters (S0, S1, S2), along with the degree of linear polarization (DoLP) and angle of polarization (AoP). The ISTA-ResUNet framework for spectral reconstruction and a resolution-preserving demosaicking algorithm for recovering high-resolution polarization images are developed. The system achieves outstanding results, including a 98.87% spectral fidelity in reconstructing a 24-color chart. Both simulations and experiments demonstrate exceptional performance in spectral and polarization image reconstruction, making this system a powerful tool for remote sensing, material inspection, and biomedical imaging.

The wing panel is subjected to complex loads during various flight missions, making real-time and accurate acquisition of its morphological features critical for evaluating structural fatigue, instability, and damage risks, as well as improving flight performance and reliability. To address this, this paper proposes a Physics-Informed Neural Network (PINN)-based wing panel shape reconstruction method, which enhances model robustness by incorporating loss functions governed by physical laws. The proposed method requires only a one-dimensional sensor array along the spanwise direction to achieve three-dimensional bending and torsional deformation reconstruction, effectively mitigating the significant shape reconstruction errors caused by sparse sensor layouts.Furthermore, a deflection function construction method based on Simpson’s integration and a based hierarchical generative neural network is developed to resolve the correlation between strain and displacement fields, thereby improving the learning efficiency of the feedback-driven PINN. Building on this framework, an experimental system for wing panel shape reconstruction using fiber Bragg grating (FBG) sensor arrays is established. Experimental results demonstrate that the proposed method achieves:•Pure bending deformation: An average relative reconstruction error of 5.86% and a root-mean-square error (RMSE) of 0.35 mm.•Pure torsional deformation: An average relative error of 7.42% and an RMSE of 0.18 mm.The findings not only contribute to aerodynamic shape optimization of wings but also provide technical support for adaptive wing morphing control.

Metasurface has drawn extensive attention due to its capability of modulating light with a high degree of freedom through ultrathin and sub-wavelength optical elements, and metalens, as one of its important applications, promises to replace the bulky refractive optics, facilitating the imaging system light-weight and compact characteristics. Besides, computer-generated holography (CGH) is of substantial interest for three-dimensional (3D) imaging technology by virtue of its ability of restoring the whole optical wave field and re-constructing the true 3D scene. Consequently, the combination of metalens and CGH holds transformative potential in enabling the miniaturization of 3D imaging systems. However, its imaging performance is subject to the aberrations and speckle noises originating from the metalens and CGH. Inspired by recent progress that computational imaging can be applied to close the gap, a novel full-color imaging system, adopting end-to-end joint optimization of metalens and CGH for high imaging quality, is proposed in this paper. The U-net based network as the pre-processing adjusts weights to make the holographic reconstruction offset imaging defects, incorporating the imaging processing into the step of generating hologram. Optimized by deep learning, the proposed imaging system is capable of full-color imaging with high fidelity in a compact form factor, envisioned to take an essential step towards the high-performance miniaturized imaging system.

In this paper, we present four denoising algorithms dedicated to the division of focal plane (DoFP) polarization image sensors, including the average filtering, median filtering, Wiener filtering and wavelet threshold denoising algorithms. Compared with the previous implementations solely based on the Gaussian noise model, this paper, for the first time, covers the non-Gaussian noises, such as the salt & pepper noise and Poisson noise. According to our extensive experimental results, the wavelet threshold denoising outperforms for suppressing the Gaussian noise; while the median filtering and the Wiener filtering outperform for suppressing the low-density salt & pepper noise and Poisson noise, respectively.

With the appearance of cameras containing newly developed polarized image sensors, which include an on-chip polarizer array placement, there is an upcoming need to characterize these cameras according to a uniform standard. These polarizer arrays consist of 2×2 matrices with four different angled polarizers, placed on top of each pixel. Therefore, this setup differs from the traditionally used multi camera and multi-polarizer setups. In terms of characterizing cameras, the EMVA 1288 standard already provides a widely established guideline on how to characterize monochrome and color digital cameras with linear photo response characteristics. The aim of this paper is to characterize a polarization camera, which takes advantage of these newly developed sensors, within the framework of the EMVA 1288 standard. Therefore, a measuring setup is constructed to be able to isolate the response characteristics of the individual pixels of the polarizer arrays. The measuring setup is built in accordance to the EMVA 1288 standard, thus consisting of a monochromatic light source, a measurement tube and a photodiode. In addition, a rotatable linear polarizing filter is being placed in front of the image sensor, to be able to control the polarization direction of the incident light. The results obtained will then be evaluated and interpreted, and proposals for an extension of the EMVA 1288 standard for the characterization of polarization cameras will be given, based on the experience gained throughout the characterization.

Water plays a critical role in Earth's climate, but monitoring its properties with autonomous underwater sampling robots remains a major challenge due to the lack of effective underwater geolocation capabilities. In this work, we present a novel underwater geolocation method using deep neural networks trained on 10 million polarization-sensitive images collected globally. Our approach achieves longitudinal accuracy of 55 km during the daytime and 1,000 km at night, at depths of up to 8 meters, irrespective of water turbidity.