The development of accurate, robust and adaptive methods for monitoring the optical parameters of fiber-optic sensors (FOS) is one of the priority tasks in the field of precision measurements, especially in the context of rapidly growing requirements for intelligent monitoring systems. This paper presents a comprehensive approach to the use of modern deep learning algorithms for analyzing and processing spectral data coming from FOS. The proposed solution is based on the use of convolutional neural networks (CNN) for the automatic extraction of informative features, as well as autoencoders for noise suppression and signal restoration. A hybrid architecture combining CNN and recurrent neural networks (RNN) was developed. The experiments conducted confirmed the effectiveness of the evaluated models. On the independent regression test set, the CNN-only model achieved a macro-averaged R2-based prediction score of 98.2% without added noise and 89.4% under high-noise conditions; on a separate temporal test sequence, the hybrid CNN + RNN model achieved 95.0% compared with 88.0% for CNN alone. The presented approach has high resistance to noise and the ability to scale to various types of FOS. At the conclusion, the prospects for the practical applications of the proposed system are discussed: the structural monitoring of buildings and structures and the automation of processes in industry and energy, with an emphasis on reliability, autonomy and integration with existing platforms.
Gold-sputtered tapered optical fiber (Au-TOF) sensors were developed for high-sensitivity refractive-index (RI) detection using aqueous glucose standards spanning 1.33–1.41 RIU (0–50% w/v). The sensors were fabricated using a reproducible workflow combining flame-brushing tapering, plasma surface preparation, and rotational magnetron sputtering, enabling azimuthally uniform gold coatings without the use of thiol or sulfur linker chemistries. Two sputtering durations (24 s and 30 s) were investigated to examine thickness-dependent plasmonic coupling and sensing performance. Optical measurements were conducted using a broadband supercontinuum source and compact spectrometer within a 20 µL microfluidic sensing chamber. Increasing glucose concentration produced a monotonic decrease in transmission intensity and a systematic red shift of the resonance minimum, consistent with enhanced evanescent-field interaction at the gold–dielectric interface. Across four independent trials, the 24 s Au-TOF sensor exhibited sensitivities up to 985 nm/RIU, while the 30 s device achieved sensitivities up to 1590 nm/RIU with strong linearity (R2 = 0.99). The enhanced sensitivity observed for the longer sputtering duration is attributed to improved gold film continuity and stronger plasmonic coupling. These results demonstrate a scalable, linker-free fabrication strategy for plasmonically enhanced tapered fiber sensors and establish the Au-TOF platform as a promising approach for label-free optical biosensing in compact microfluidic environments.
Fringe projection profilometry (FPP) is widely used for high-precision three-dimensional measurement, but its performance is severely degraded when measuring non-Lambertian surfaces with complex reflectance. In such cases, fringe saturation and low modulation lead to unreliable phase values and missing reconstructed data. To address this problem, this paper proposes a photometric feature-guided phase inpainting method for high-dynamic-range FPP. Within the proposed FPP with multi-illumination framework, fringe images are used for phase calculation, while multi-illumination images provide additional photometric cues for phase inpainting. A photometric feature-guided phase inpainting network is designed to extract photometric features from multi-illumination images and fuse them with the damaged phase map. The network is trained on a large-scale synthetic dataset and validated using a prototype system. In standard copper-sphere measurements, PF-PINet increases the average reliable reconstruction ratio from 81.85% to 84.31% while maintaining comparable reconstruction accuracy, indicating that the proposed method can improve measurement completeness for non-Lambertian surfaces.
A plasmonic metamaterial absorber (MA) exhibits unique capabilities of electric field enhancement and spectral modulation and plays a critical role in biosensing applications. In this work, a narrowband MA structure based on a plasmonic square nanoring is designed and numerically investigated. The proposed MA exhibits a high narrowband absorption of 99.77% at a 1216 nm wavelength with a 10 nm full width at half maximum (FWHM). Furthermore, the MA has sensitive absorption properties to the refractive index (RI) of the environment and shows a sensitivity of 186 nm/RIU with a stable value of quality factor around 120 in the RI ranging from 1.0 to 1.4. The sensing and recognition ability of the MA on actual biological cells is also numerically verified. Meanwhile, the proposed MA exhibits excellent capability of fluorescence amplification, achieving 3438–fold fluorescence enhancement at a 1216 nm wavelength with a high directivity of ~590. These properties make the proposed MA structure an excellent candidate for application in an optical sensing system.
Interferometric fiber-optic hydrophones are widely deployed in passive underwater acoustic detection systems due to their high sensitivity, broad dynamic range, and immunity to electromagnetic interference. However, their adoption in active acoustic systems has been largely constrained by the limited operational bandwidth of conventional architectures. Through coupled acoustic–structural simulations, this work identifies the operational bandwidth bottleneck as arising from low-order mechanical resonance of high-modulus mandrel structures combined with acoustic cavity resonance and near-field scattering within the enclosed cylindrical geometry. To address these limitations, an optimized push–pull mandrel structure, featuring a miniaturized profile of φ16 mm × 16 mm and a low-acoustic-impedance material, is implemented to eliminate mechanical resonance within the operational bandwidth. Additionally, an integrated acoustic metamaterial liner is incorporated into the inner tube to mitigate scattering and reverberation effects. Experimental characterization of the prototype demonstrates an average phase sensitivity of −132 dB re 1 rad/μPa, with fluctuations below ±1.5 dB across 20 Hz to 31.5 kHz. Utilizing a custom-built demodulation system, the hydrophone achieves a minimum detectable pressure of 42 dB re 1 µPa/√Hz (126 µPa/√Hz) at 20 Hz and 27 dB re 1 µPa/√Hz (22 µPa/√Hz) above 2 kHz, remaining over 28 dB below the deep-sea state zero ambient noise floor at 20 Hz, and hence confirms its passive acoustic monitoring capability. In the high-frequency regime above 2 kHz, the hydrophone maintains ±1.5 dB sensitivity flatness with a horizontal directivity variation of only ±1.51 dB at 30 kHz, providing a robust platform for broadband active acoustic applications, including active sonar, underwater acoustic imaging and acoustic communications. These results establish a viable pathway toward next-generation dual-mode underwater acoustic systems requiring both high-fidelity passive listening and broadband active detection.
Optical bistability (OB) is a fundamental nonlinear effect enabling all-optical switching, logic, memory, and sensing, yet its practical implementation is often hindered by high power thresholds and limited tunability. Here, we theoretically demonstrate low-threshold and widely tunable OB in a Fabry-Pérot (FP) cavity integrated with a Weyl semimetal (WSM) layer. By combining the strong local-field enhancement of the FP cavity with the large third-order nonlinear conductivity of the WSM, we achieve pronounced bistable hysteresis loops in both the transmitted field amplitude and transmittance at terahertz frequencies, with incident field switching thresholds on the order of 106 V/m. The bistability thresholds and hysteresis width can be flexibly controlled by tuning the WSM Fermi energy, the position of the WSM layer inside the cavity, and the mirror transmittance. Moreover, exploiting the strong dependence of the down-switching threshold on small changes in the refractive index of the cavity medium, we propose a terahertz refractive-index sensor with a maximum sensitivity of 296.8 GW·m−2·RIU−1. These results provide a theoretical basis for tunable nonlinear photonic devices and threshold-based terahertz refractive-index sensors using topological semimetals.
The complex dispersion and modal sensitivity of an axisymmetric transverse magnetic surface wave supported by a dielectric-coated perfectly conducting cylinder are investigated. Starting from Maxwell’s equations, the boundary-value problem is reduced to a nonlinear complex dispersion equation for the longitudinal propagation constant β. A numerical framework combining zero-level localization of the real and imaginary parts of the dispersion function, nonlinear root refinement, numerical clustering, and adaptive continuation in the complex coating permittivity is used to identify and track a selected spectral branch. One- and two-parameter computations characterize the mapping ε↦β(ε) over prescribed subsets of the complex-permittivity plane. At fixed ℑε, increasing ℜε increases ℜβ and decreases ℑβ, whereas at fixed ℜε, increasing ℑε increases both components of β in the investigated parameter range. The rectangular-grid, concentric-circle, and radial-beam experiments show that the spectral response is smooth on the considered parameter sets but non-affine, coupled, and direction-dependent. The corresponding longitudinal electric field is reconstructed, normalized, and phase-aligned along the tracked branch. Difference fields, radial localization measures, a global modal distance, and a normalized correlation coefficient show that the same qualitative radial TM mode is retained throughout the sampled parameter domain, while its propagation constant and spatial localization vary continuously with the complex coating permittivity.
Polarization–spectral imaging information has theoretical advantages for camouflage-net recognition. However, for specific observation targets, the selection of effective spectral bands and polarization parameters, which are critical issues in engineering applications, remains unclear. Using a newly developed hyperspectral polarization imaging prototype, we conducted systematic outdoor observations of a grassland camouflage net placed over green grass at multiple viewing angles and different times of day. After evaluating and confirming the net’s camouflage effectiveness, we analyzed polarization-parameter spectral images of both the net and background under varying observation and solar altitude angles. The Fisher criterion was used to identify discriminative polarization parameters and spectral ranges. Effective bands were determined as S1 at 675–696 nm, S2 at 673–677 nm, S3 at 607–616 nm and 625–700 nm, DoP at 676–700 nm, and DoLP at 668–680 nm. Grayscale images in these bands further verified the separability of the camouflage net from grass. These results provide experimental evidence and wavelength-parameter references for engineering applications of polarization–spectral imaging in camouflage-net detection.
Neuromorphic imaging sensors (event cameras) offer a promising paradigm for computational imaging and human pose estimation (HPE) under extreme illumination conditions. Nevertheless, dark-scene background activity originating from photodiode dark current and circuit thermal noise, together with hot-pixel noise, severely corrupts event streams and impedes reliable HPE in low-light scenarios. To address this issue, we propose an adaptive event denoising framework built upon a spatiotemporal Gaussian-weighted neighborhood model with a dynamic thresholding mechanism. It can effectively suppress background activity and hot-pixel noise while preserving edge and motion details critical for pose estimation. Leveraging this denoising front-end, we construct a complete dark-scene neuromorphic HPE pipeline by transferring the pre-trained MediaPipe model onto event-based time-surfaces. Quantitative and qualitative evaluations on public and self-collected datasets demonstrate that our approach outperforms state-of-the-art denoising methods with an improvement of over 20% on public benchmarks and over 30% on self-collected dark-scene data. We expect our work to pave the way toward reliable dark-scene human–robot interaction through robust neuromorphic pose estimation.
Lensless coded ptychography (CP) acquires multiple intensity measurements by translating either the object or the coded sensor, and reconstructs high-resolution, large field-of-view images via iterative phase retrieval. However, conventional CP suffers from slow reconstruction due to two limitations. First, periodic uniform scanning requires dense measurements to maintain sufficient measurement diversity, resulting in a large number of raw measurements and a high computational burden for iterative phase retrieval. Second, random initialization leads to slow convergence and increases the risk of stagnation in local minima. To address these limitations, we propose a joint optimization of the sampling strategy and initialization for accelerated reconstruction. Specifically, a continuous non-uniform scanning strategy preserves measurement diversity while reducing the number of required measurements. In addition, a low-resolution regularized ptychographic iterative engine (rPIE)-based initialization provides a more accurate starting point and accelerates the convergence of iterative phase retrieval. Both simulations and experiments demonstrate that the proposed approach achieves approximately three-fold faster convergence while maintaining high reconstruction fidelity. The proposed approach offers an effective solution for high-throughput imaging applications, including digital pathology and label-free quantitative phase imaging.
A simple scheme for generating a broadened and flattened optical frequency comb (OFC) based on a directly modulated distributed feedback (DFB) semiconductor laser cascaded with electro-optic modulators is proposed. The initial OFC is generated by a gain-switched DFB semiconductor laser. The number of comb lines is further expanded by cascading a phase modulator and the flatness performance is optimized by adjusting the power of the radio frequency (RF) and bias voltage of the cascaded Mach–Zehnder modulator. Moreover, the spacing of comb lines can be flexibly tuned by varying the frequency of the RF signal, and an OFC with seven spectral lines within a 4.72 dB power variation is demonstrated. The 17-line OFC with a frequency spacing of 12.5 GHz and a flatness of 3 dB is further applied to an optical transmission system, which enables error-free transmission of a non-return-to-zero data signal. The proposed scheme is expected to find applications in flexible-grid optical transmission systems.
Free-space optical (FSO) communication provides high-capacity wireless transmission but suffers from reduced optical coupling efficiency when compact Cassegrain telescopes are employed because the secondary mirror blocks the central portion of the incident beam. This paper proposes a vortex beam-assisted FSO communication system that combines aperture-matched optical coupling with machine learning-based signal recovery. Unlike a conventional Gaussian beam, the annular intensity distribution of a Laguerre–Gaussian vortex beam is matched to the unobstructed annular aperture of a centrally obscured Cassegrain telescope, thereby reducing obstruction-induced optical loss. The coupling characteristics are analyzed using an annular aperture overlap model and experimentally validated in a 100 m free-space optical link employing Cassegrain transmitter and receiver front-ends. The mean measured telescope-output power is increased by more than 30% over the Gaussian reference. To overcome the system-induced signal aliasing, Transformer and long short-term memory (LSTM) equalizers are optimized and applied. Both models substantially outperform optimized threshold detection, while the LSTM achieves the lowest observed error rate with markedly fewer multiply–accumulate operations than the Transformer. These results show that aperture-matched optical coupling and computationally efficient sequence equalization, such as LSTM is a crucial component of compact, high-performance telescope-assisted FSO systems.
Light fidelity (LiFi) is progressively evolving as a highly promising communication technology because of its unique benefits, available spectrum, low implementation costs, and adaptive beamforming capabilities. Despite their advantages, existing LiFi networks remain constrained by limited data rates, coverage area, and information security in practical environments. Therefore, a high-speed, high-capacity, and secure quantum key distribution (QKD)-assisted integrated multi-wavelengths (450/532/620 nm) LiFi system using mode division multiplexing (MDM) is proposed. The results demonstrate that the proposed system achieves maximum transmission distances of 20.5–22 m and 19–22 m using different Laguerre–Gaussian (LG) and Hermite–Gaussian (HG) mode indices {[0,0], [0,10], [0,20], [0,30]}, at an aggregate data rate of 40 Gbps. Furthermore, the minimum acceptable transmitter angles of 30–90° for irradiance angles of 20–80° are required to maintain the target bit error rate (BER) of 10−9. The minimum photodetector detection areas required at transmission distances of 20–30 m are 1–2 cm2 at the minimum BER limit. Moreover, the proposed system exhibits optimum performance, achieving an optical loss of −39.47 dB, −49.03 dBm received power, and 45.39 dB signal-to-noise ratio for 1–10 photons/pulse. Compared with existing studies, the proposed system demonstrates enhanced overall performance across various communication metrics.
By analyzing the semantic information of Direct Current (DC, the zeroth-order spherical harmonic coefficient) gradients during 3D Gaussian Splatting (3DGS) optimization, this paper achieves unsupervised state classification in scenes with discrete appearance states under the proposed State-Discovery Gaussian Splatting (SD-GS) framework via SVD dimensionality reduction and K-means clustering. To improve the stability of the clustering results, an appearance-difference-weighted refinement mechanism is further proposed to confirm high-confidence labels. To address the difficulty of distinguishing similar states when the number of states exceeds two, a sequential peeling strategy is proposed that decomposes a multi-class partition into several two-class separations. On four real-world scene datasets, SD-GS achieves 100% classification accuracy with reconstruction quality of 31.98–38.83 dB PSNR. Ablation studies validate the effectiveness of the gradient direction mode and the SVD dimensionality reduction strategy.
We obtain exact solutions for high-order hybrid cladding modes of standard fibers, paying special attention to the case in which two hybrid cladding modes have very close propagation constants (crossover points). We calculate the mode fields, dispersion, and polarization distribution of cladding modes at crossover. We discuss the applicability of the linearly polarized modes approximation for calculating modes in this case. We show that, at the crossover points, the modes are not standard HE and EH hybrid modes but radial and azimuthal modes with field distributions resembling those of TE and TM modes. We analyze the dispersion of hybrid modes with azimuthal number equal to 1 and find crossover points in the range 0.6–1.7 μm for fibers with various V-numbers. For standard fibers, the first 23 hybrid modes have crossovers at wavelengths below 1.2 μm. Accounting for the new hybrid modes reveals the splitting of resonances in long-period fiber gratings. A linear combination of crossover modes can be used to form approximate HE and EH modes with uniform linear polarization for the HE mode and a magnetic-dipole-like field for the EH mode.
Land-based hyperspectral imaging provides high spatial and spectral resolution for detecting camouflaged targets, but practical deployment remains limited by strong target background spectral similarity, scarce annotated hyperspectral samples, and the computational cost of full-band processing. To address these issues, this paper proposes HCTDNet (Hyperspectral Camouflaged Target Detection Network), a land-based hyperspectral image analysis framework. The method first employs band extraction for data dimensionality reduction, compressing multi-channel hyperspectral images into 3-channel virtual RGB representations, which reduces spectral redundancy while preliminarily enhancing camouflaged target saliency. A pre-trained RGB camouflaged target detector is then adopted as the backbone model, with its parameters frozen to maintain stability, while trainable modality-specific prompts are learned to improve training efficiency. Finally, model fine-tuning is performed using a self-constructed camouflaged target dataset to enhance robustness in detecting camouflaged targets within virtual RGB images. During inference, preprocessed hyperspectral images are fed into the model to generate detection results for camouflaged target regions. The experiments performed on our self-collected land-based hyperspectral dataset with camouflaged targets reveal that HCTDNet achieves superior detection performance compared with seven classical hyperspectral target detection methods while maintaining an average inference speed of approximately 16 FPS. The proposed framework provides an efficient and near-real-time applicable solution for land-based hyperspectral camouflaged target detection, showing significant practical potential.
Visible light communication (VLC) has emerged as a transformative optical wireless technology for sixth-generation (6G) networks, offering license-free spectrum access, inherent electromagnetic-interference immunity, high spatial confinement, and the unique ability to combine high-speed wireless connectivity with solid-state lighting infrastructure. However, the transition from conventional VLC links to practical 6G optical wireless systems requires far more than advanced modulation and signal processing. Future VLC performance will be strongly determined by the co-design of photonic front-ends, including high-speed transmitters, spectrally engineered emitters, reconfigurable optical interfaces, intelligent receivers, and energy-autonomous detection units. This article provides a comprehensive, device-centered review of photonic hardware and artificial intelligence (AI) enablers for next-generation 6G VLC systems. Particular attention is given to micro-LEDs, laser diodes, color-conversion materials, including perovskite quantum dots, advanced photodetectors, imaging receivers, wavelength-shifting fiber receivers, solar-cell-based receivers, optical reconfigurable intelligent surfaces (RISs), metasurfaces, beam-steering components, and optical wireless power transfer. This review discusses how AI can support inverse photonic design, transmitter and receiver calibration, nonlinear impairment mitigation, channel-aware beam control, and energy-aware resource management. Unlike broader VLC surveys that mainly emphasize network architecture, this article provides a device-centered perspective on AI-enabled photonic integration for 6G VLC, supported by a comprehensive survey of recent experimental demonstrations. Key challenges related to bandwidth, optical efficiency, receiver field of view, mobility, safety, standardization, and practical deployment are summarized, followed by a research roadmap for 2025–2032.
Optical coherence tomography (OCT) and photoacoustic imaging (PAI) provide complementary structural and absorption contrasts but require different coupling conditions: 1310 nm swept-source OCT is attenuated by water, whereas PAI requires acoustic coupling. We developed a large-field dual-modal imaging system combining temporal medium separation with hardware-based coordinate locking. The OCT head, linear-array ultrasound transducer, and photoacoustic excitation fiber bundle were mounted on a rigid common platform, and a one-time calibration established a two-dimensional affine transformation between the modality coordinate systems. OCT was acquired in air and PAI in deionized water within a common large-field coordinate range. In five paired air–water measurements with an approximately 23 mm water path, the displayed OCT peak level decreased from 98.4 ± 1.5 dB in air to 79.4 ± 1.8 dB in water, corresponding to a mean reduction of 19.0 ± 1.4 dB. Quantitative registration was evaluated using a 5 × 5 dual-modal landmark phantom, with nine landmarks used for affine calibration and 16 excluded landmarks reserved for independent validation. The mean two-dimensional validation error was 0.235 ± 0.128 mm, with an RMSE of 0.266 mm and a maximum error of 0.446 mm. Five additional medium-switching cycles performed without recalibration yielded an overall registration error of 0.369 ± 0.163 mm across 80 validation measurements. PA spatial resolution was further characterized using six thin hair targets, yielding lateral and axial FWHM values of 0.342 ± 0.069 mm and 0.394 ± 0.073 mm, respectively. These results demonstrate reproducible two-dimensional en face OCT–PA coordinate mapping under modality-specific coupling conditions and support the proposed workflow as a phantom-based technical validation for large-field multimodal imaging.
Underground cables are critical infrastructure for electrical power and communication transmission, and their reliable operation is of paramount importance to urban public safety. Although Distributed Acoustic Sensing (DAS) enables wide-range, continuous, and real-time monitoring, traditional DAS signal processing methods suffer from poor intrusion discrimination and weak anti-interference capability. To address these limitations, we propose a dual-branch network based on multi-domain feature fusion, integrating a Multilayer Perceptron, a Long Short-Term Memory network (LSTM), and an attention mechanism. Vibration signals corresponding to four representative high-risk intrusion events were acquired through controlled field experiments, and a standardized, category-balanced dataset was constructed accordingly. Time-domain, frequency-domain and joint time-frequency features were extracted and mapped through a time-frequency weighting transformation to form one branch of the network, while the parallel branch employed an LSTM to capture long-range temporal dependencies. A multi-head attention mechanism enables deep adaptive fusion of two types of modal information and overcomes the limitations of conventional simple feature concatenation. Comparative experiments against KNN, 1D-CNN and LSTM baselines demonstrate that the proposed model achieves a test accuracy of 98.89%, outperforming all reference methods. Ablation studies further validate the necessity and effectiveness of each constituent module within the proposed architecture. The results indicate that this approach provides reliable support for DAS-based online monitoring of power cables against external damage.
Hyperspectral unmixing (HU) requires effective modeling of spectral–spatial information and local–global feature interactions to achieve accurate abundance estimation and endmember extraction. Although Transformer-based HU methods are effective in capturing long-range dependencies, they often neglect the intrinsic spectral consistency of hyperspectral data and do not fully exploit the global spectral–spatial correlations in hyperspectral image cubes. To address these issues, this paper proposes a Spectral Consistency–guided Transformer with Volumetric Linear Self-Attention (SCVolFormer) for hyperspectral unmixing. A Spectral Consistency Block (SCB) is introduced to preserve consistency across adjacent spectral bands and produce physically meaningful feature representations. A spectral grouping strategy is further adopted to partition the high-dimensional spectrum into locally continuous subspaces, reducing computational cost. In addition, a shared-weight Transformer encoder with Volumetric Linear Self-Attention (VolLSA) is designed to model interactions between the spectral and spatial dimensions and capture long-range dependencies within hyperspectral image cubes. A decoder is then used to estimate abundance maps and reconstruct hyperspectral images. Experiments on one synthetic dataset and three real hyperspectral datasets demonstrate that SCVolFormer outperforms state-of-the-art methods in abundance estimation and endmember extraction, confirming the effectiveness of spectral consistency guidance and volumetric attention modeling for hyperspectral unmixing.
