High-precision aspheric surface testing follows the null-test principle, in which the computer-generated hologram (CGH) is an indispensable compensating element for aspheric null testing. In aspheric null interferometry based on a null CGH and a Fizeau interferometer, the measurement accuracy is significantly affected by fabrication errors of the CGH. Absolute testing achieves higher measurement accuracy than conventional aspheric null interferometry by separating the fabrication errors of the CGH. In this paper, an absolute testing method based on transmissive Twin-CGH is proposed for rotationally symmetric convex aspheric surfaces. By fixing the position of the Twin-CGH during absolute testing, the wavefront error introduced by the Twin-CGH and transmission sphere (TS) can be separated from the surface figure error of the aspheric mirror, thereby avoiding errors caused by CGH pose adjustment. Compared with the conventional transmission-type Twin-CGH method, this approach eliminates the need to precisely position a standard spherical reflector at the conjugate focus of the spherical wave. It bypasses a critical bottleneck: the conflict between conjugate positioning requirements in confined optical paths and CGH manufacturability constraints. Instead, it leverages high-precision lithography techniques such as electron beam lithography and laser direct writing to keep the pattern distortion of the Twin-CGH at a low level, offering high feasibility. To verify the validity and reliability of the proposed method, a Twin-CGH and a null CGH are fabricated separately. The proposed method was validated through two independent approaches: absolute testing using the Twin-CGH, and error separation via the N-position method with a null CGH. The results obtained by the two methods verify each other, which fully demonstrates the correctness and reliability of the proposed testing method. It is pointed out that the practical design of the Twin-CGH needs to comprehensively consider the spatial distribution of the 15 diffraction combinations listed in Table 3 at the aperture
Femtosecond laser direct writing has emerged as a powerful and versatile fabrication technology for realizing advanced optical fibre devices with high precision, excellent material compatibility, and three-dimensional processing flexibility. By utilizing nonlinear light–matter interaction, fs-laser direct writing enables localized refractive index modification and micro/nano-processing to optical fibre, making it particularly attractive for fabricating fibre Bragg gratings, interferometric devices, and fibre-integrated photonic components. This review summarizes the fundamental mechanisms of fs-laser-induced material modification and discusses major direct-writing techniques, including point-by-point, line-by-line, and plane-by-plane inscription methods. Recent advances in beam-shaping approaches for improving the spectral performance of fibre Bragg gratings are also reviewed. In addition, the fabrication and applications of complex grating structures, including apodised, chirped, tilted, helical, and eccentric FBGs, as well as grating fabrication in specialty optical fibres are discussed together with their roles in sensing, optical filtering, and fibre laser systems. The review further highlights the use of fs-laser direct writing for fabricating compact interferometric devices, for customizing optical components to be integrated in optical fibre through non-deformation, and additive processes. Overall, the flexibility of fs-laser fabrication enables the realization of customized optical fibre devices from laboratory demonstrations to practical applications.
A Brillouin scattering spectrum noise model is proposed to improve the accuracy of numerically generated noisy spontaneous Brillouin scattering spectrum. Firstly, the Brillouin gain noises and Brillouin parameters are obtained by analyzing a large number of Brillouin spectra with different incident light pulse widths and signal-to-noise ratios. It is found that the noise follows a Gaussian distribution with zero mean. Then, the relationship between variance of Brillouin gain noise and Brillouin gain is derived. Further, the noise model is proposed. Based on the proposed noise model, typical noise model and the measured spectra, the parameters of the noise models are determined. According to the relevant parameters obtained from the measured spectra, the corresponding Brillouin spectra based on the proposed noise model and typical noise model are numerically generated. Furthermore, Brillouin frequency shift (BFS) is calculated by the single slope-assisted technique according to the measured spectra and the numerically generated spectra. The difference in the calculated BFS errors between the typical noise model and the measured spectra is 17.85 MHz. Meanwhile, for the proposed noise model, the difference is only 0.57 MHz. Furthermore, the proposed model is validated through BFS fluctuation analysis using the pseudo-Voigt and Lorentzian fitting methods, as well as the ANN-based BFS estimation. The results reveal that it can accurately characterize Brillouin gain noise and which can be used to improve BFS extraction accuracy.
We investigate terahertz (THz) single-pixel imaging under spatially non-uniform illumination using a quantum cascade laser source with structured beam profiles. The reconstruction problem is formulated as a self-calibrated framework. The unknown illumination field is estimated directly from single-pixel measurements. It is then incorporated into the sensing model for object reconstruction. The approach is evaluated with computational ghost imaging and with compressive sensing methods, including OMP and TVAL3. All quality metrics are reported as distributions over random subsets of the recorded measurements, and the method comparisons are supported by paired significance tests. We find that the benefit of self-calibration depends on the degree of illumination inhomogeneity. Under strongly non-uniform illumination, the self-calibrated TVAL3 method provides the highest reconstruction quality and significantly outperforms all single-stage methods. Under weakly non-uniform illumination, the error of the beam estimate outweighs the gain from the illumination correction, and single-stage reconstruction remains preferable. The transition between the two regimes is characterized by the intensity coefficient of variation within the object region. The proposed framework enables THz imaging without beam homogenization optics, which is particularly relevant for compact THz systems based on quantum cascade lasers.
Achieving ultra-broadband and highly efficient achromatic metalenses in the near-infrared (NIR) and shortwave infrared (SWIR) regimes remains a formidable challenge due to the inherent absorption and refractive-index dispersion of conventional nanophotonic materials. In this numerical study, we propose an ultra-broadband, polarization-insensitive achromatic metalens operating across a continuous spectral range from 700 nm to 2500 nm, corresponding to an absolute bandwidth of 1800 nm, utilizing aluminum nitride (AlN) as a dielectric material platform. By systematically engineering a diverse meta-atom library and optimizing its spatial distribution through rigorous coupled-wave analysis (RCWA), we successfully achieve precise phase and phase-dispersion compensation. Full-wave finite-difference time-domain (FDTD) simulations demonstrate that the proposed design provides a record-breaking achromatic bandwidth with a numerical aperture of 0.322. The simulated metalens exhibits high-performance focusing, featuring an exceptional average transmittance of 96.8% and an outstanding average focusing efficiency of 78.9%, peaking at 87.3%. Furthermore, the device maintains near-diffraction-limited focusing across the entire target spectrum, evidenced by an average Strehl ratio of 0.864 and remarkable focal stability. Additionally, the off-axis focusing performance under oblique incidence demonstrates a practical operational field of view of approximately 20°. These results establish a new benchmark in the literature regarding both operational bandwidth and optical efficiency, paving the way for advanced, highly compact, and multi-spectral imaging and sensing systems.
Passive autofocus systems are crucial in microscopic imaging. However, traditional search algorithms are highly susceptible to falling into local extrema when confronted with high magnifications, shallow depths of field, or complex texture perturbations, resulting in autofocus failures. To address this issue, an improved hill-climbing search algorithm assisted by a frequency-domain energy factor is proposed in this paper. Starting from the physical mechanism of a defocused system, the algorithm utilizes the low-pass filtering characteristics of the system on high-frequency signals to construct an extremum discrimination factor, Ek, based on significant low- and mid-frequency energy. During the coarse-search stage of a predefined forward scan, Ekis evaluated at suspected sharpness peaks to reject pre-focus false candidates and verify the principal-peak candidate before fine localization, thereby providing a directional physical-consistency check. Combined with an objective-dependent “coarse-medium-fine” three-stage preset step-size switching strategy, the proposed algorithm improves robustness while maintaining rapid convergence. Systematic experimental validations were conducted on various biological samples, including BPAE cells and live MCF-7 cells, under 10× to 60× objectives. The results indicate that the proposed algorithm maintains a near 100% autofocus success rate across all tested magnifications and reduces motor step counts by approximately 60% compared to the rule-based autofocus method. The reduction in motor steps may help decrease mechanical actuation, system energy consumption, and fluorescence excitation exposure.
The method described in this paper is related to scanning profilometry and aims to improve measurement accuracy. It is shown that when precision optical sensors for on-the-fly measurement of the machined surface profile are integrated into production equipment, measurement accuracy is often limited not by the intrinsic resolution of the sensor but by errors in the scanning mechanics.A software method is proposed to compensate for these errors (mechanical errors, thermal drifts, &c.) when measuring surface form under such conditions. The approach is based on the use of a reference planar area within the scan region and on explicit accounting for the sequential, line-by-line structure of the measurements. Errors are separated by characteristic temporal scales into slow drifts, quasi-stationary within a line, and fast reproducible errors that repeat from line to line. Compensation is performed using data from a single measurement without referring to external standards or additional calibration procedures.The method’s viability is demonstrated experimentally during inspection of optical elements, including quantitative validation on a certified step-height reference. Systematic errors are shown to be suppressed from the micron level to a magnitude limited by the resolution of the confocal sensor (≈50 nm) and flatness of reference area surface.
In the field of high-performance manufacturing, rapid and accurate acquisition of the topography of aluminum alloy machined surfaces is crucial for evaluating its in-service performance. Among various non-destructive inspection methods, the Shape from Shading (SFS) method has attracted widespread attention due to its simple equipment and high computational efficiency. However, traditional iterative SFS methods are severely limited by the inherent concave-convex ambiguity problem, which easily leads to morphological distortion and depth deviation. Moreover, existing improvement schemes generally suffer from additional hardware dependencies, complex system structures or poor generalization capability. Therefore, this paper proposes an analytical SFS reconstruction method for aluminum alloy machined surfaces based on concave-convex discrimination via image definition. This study analyzes the geometric relationship between the machined surface and imaging characteristics, establishes a mapping model between surface gradients and image grayscales, and reveals the intrinsic correlation mechanism between image definition and surface concave-convex features. On this basis, a concave-convex discrimination strategy is constructed, and an analytical model is further developed for aluminum alloy machined surfaces. Experimental results show that this method can rapidly and accurately identify the concave-convex features of aluminum alloy machined surfaces, exhibiting superior performance in terms of reconstruction accuracy and computational efficiency. It provides a new approach for the online quality inspection of high-performance components.
Absolute testing in interferometry requires precise separation of reference and test surface errors. Traditional methods, such as Zernike polynomial decomposition and influence function-based approaches, often face a trade-off between accuracy, computational efficiency, and flexibility. This paper proposes a novel iterative reconstruction algorithm that employs an adaptive relaxation factor. By dynamically adjusting the relaxation parameter during iteration, the method strikes a favorable balance between convergence rate and stability, i.e., at the early stage, it ensures stable and rapid convergence, while near approaching the optimum, it introduces controlled divergence to escape local minima and seeks a superior solution. This enables high-precision, high-resolution reconstruction well suited for large-data-volume calibration scenarios. Comprehensive simulations demonstrate that, under controlled noise conditions, the proposed algorithm achieves reconstruction residuals approaching the simulated noise level (∼0.1 nm RMS), and exhibits a favorable balance between convergence speed and stability compared to fixed-parameter methods. The practical noise floor of the experimental interferometer is further characterized through repeatability measurements, yielding a pixel-wise standard deviation RMS of 0.11 nm (1σ). Experimental validation, including a conclusive ion-beam figuring test conducted with a phase-shifting interferometer, confirms the practical efficacy of the method. The results indicate that the proposed adaptive relaxation method provides a robust, efficient, and high-precision solution for absolute surface metrology, particularly suitable for high-resolution data processing.
Terahertz defect detection in composites remains difficult when labelled samples are scarce, class distributions are imbalanced, and model predictions lack physical consistency. We propose a Physics-informed Constrained Inception-TCN-Attention Network (PI-ITCNANet) for small-sample terahertz signal analysis. The model supports defect classification, continuous depth estimation, and two- and three-dimensional defect imaging. The proposed model extracts multi-scale echo features using Inception modules. It uses dilated convolutions in the TCN module to capture long-range temporal dependencies among surface, defect, and bottom echoes. Channel and temporal attention mechanisms are further introduced to enhance defect-related responses. During training, multiple physical constraints are incorporated. These constraints encourage the predictions to satisfy both data supervision and the physical relationships of terahertz wave propagation. Experimental results show that PI-ITCNANet achieves Macro-F1 score of approximately 98.67%, depth MAE of 0.170 mm, and Dice coefficient of 0.957 for defect regions. The two- and three-dimensional reconstruction results further demonstrate that the proposed method can effectively recover defect locations, depth levels, and spatial distributions. These results indicate its potential as an intelligent recognition and imaging approach for terahertz nondestructive testing of composite materials.
Although phase-shifting fringe projection profilometry (FPP) enables high-precision, non-contact 3D measurements, its accuracy is severely degraded by highly reflective surfaces (e.g., metals and smooth coatings) that induce local overexposure, saturation clipping, and low-contrast artifacts. These distortions corrupt phase retrieval and subsequent 3D reconstructions. To resolve these limitations, we propose FFG-Net, a physics-guided neural network designed for robust, single-frame fringe pattern recovery. Unlike conventional spatial-domain image restoration methods, our approach directly incorporates a frequency-domain prior into the deep feature modeling process, leveraging the fundamental physical periodicity and directionality of projected fringes. Specifically, a lightweight frequency selection block (FSB) is embedded within the bottleneck layer of a symmetric U-Net framework to adaptively recalibrate deep spatial features via frequency-domain amplitude responses. This block derives channel-wise descriptors from the amplitude spectra and adaptively recalibrates the bottleneck feature channels, emphasizing channels associated with coherent fringe structures while attenuating those dominated by reflection-induced disturbances. Operating end-to-end on single degraded frames, the network encourages consistency with diffuse reference fringes using ideal diffuse fringes as supervisory labels. The restored fringe patterns can be directly integrated into standard multi-frequency phase-shifting and temporal phase unwrapping pipelines without requiring hardware modifications or extra projection sequences. Extensive evaluations on both synthetic and empirical datasets demonstrate that our method substantially improves 2D fringe quality and 3D reconstruction completeness, improving valid point-cloud density. FFG-Net provides an efficient software-based fringe restoration strategy for high-reflection FPP measurement in industrial metrology.
Speckle scrambling fundamentally limits the practical applications of multimode fibers (MMFs). Current pure-digital or pure-optical methods struggle to balance computational overhead with reconstruction fidelity. To address this challenge, we propose a lightweight synergistic reconstruction framework that incorporates a numerically simulated, physics-based multi-plane light conversion (MPLC) front-end as a diffractive preprocessing stage prior to digital decoding. Crucially, the simulated MPLC preprocessor remaps the spatially disordered speckle field into low-complexity "pseudo-speckle" priors via cascaded phase modulations and intensity extraction. Numerical evaluations demonstrate that this framework compresses the total system parameters to 4.96 M (approximately 11% of a standard Pix2Pix model) while achieving a Structural Similarity Index (SSIM) of 0.7990±0.003on the MNIST dataset. Compared to purely digital methods, this architecture achieves a favorable parameter-fidelity trade-off: reducing complexity by 89% with moderate performance sacrifice, while demonstrating moderate robustness under dynamic fiber perturbations (SSIM 0.7104±0.0020on perturbed MNIST-D). Ultimately, this simulated framework establishes a pragmatic numerical baseline for future hardware implementations in edge computing and endoscopy.
