Digital holography (DH) has been widely utilized for in-situ observation of plankton due to its non-destructive nature, high resolution, and large depth of field. However, current in-situ DH techniques still face significant challenges, including slow reconstruction speed, poor image quality, and limited reliability. To overcome these challenges, we propose a multi-autofocusing off-axis DH approach for fast and robust multi-object plankton imaging. This method incorporates an axial sliding minimum fusion strategy to rapidly integrate target field slices, enabling dynamic localization of all regions of interest (ROIs). A hybrid metric function is then utilized to precisely focus on all objects within the ROIs, adaptively determining their optimal focal distances. Consequently, all plankton in the hologram can be reconstructed with high fidelity and stability. The proposed method significantly enhances reconstruction accuracy and robustness, making it well-suited for dynamic and complex underwater imaging environments. Experimental results confirm that our approach substantially improves reconstruction speed, imaging accuracy, and stability compared to conventional methods.

The point diffraction interferometer (PDI) is a promising quantitative phase imaging (QPI) method, which has the advantages of compactness and stability. However, the field-of-view (FOV) of PDI is always compromised between the size of the sensor and the magnification. To solve this problem, a PDI with doubled FOV is set up by a grating placed outside the Fourier plane in a 4f system, which has a simple optical setup and larger FOV without decreasing the magnification. First, a 4f system is built up by two Lens. Then, a grating is placed outside the Fourier plane of the 4f system, while a hole array is placed exactly at the Fourier plane. The grating diffracts the object beam into several duplicates with relative offsets along its periodicity, each of which carries a different region of the object. The hole array comprises one pinhole and two large holes. One of +1 diffraction orders is low-pass filtering by the pinhole to form the reference beam, while the other one of +1 diffraction orders and 0th diffraction order pass through the large holes and act as the object beams with different FOV. The image sensor is placed at an overlapping area of two FOVs, which enables two distinct regions of the object to be captured simultaneously in a single shot. Moreover, induced by the different angles between the reference beam and the object beams, object beams with different FOVs have different spatial carrier frequencies in the multiplexed interferogram. To avoid crosstalk between the object beams, two object beams are modulated into orthogonal polarization states to avoid interference. The validity and feasibility of this PDI are verified by conducting experiments on a 1951USAF resolution plate, a bee wing, and onion epidermal cells. The experimental results show that this proposed PDI can double FOV without sacrificing image quality, which demonstrates various future applications in microscopic imaging and optical metrology.

A cascaded Fourier transform reconstruction algorithm is proposed for dual-wavelength off-axis digital holography with isotropic carrier frequencies. The algorithm begins by reconstructing the +1st order complex amplitudes from the single-shot hologram at two wavelengths. These complex fields are then conjugately multiplied to generate a synthetic-wavelength hologram. By applying a subsequent Fourier transform to this hologram, the unwrapped phase difference is directly reconstructed. Unlike conventional algorithms, the proposed algorithm bypasses the separate reconstruction of single wavelength phases and subsequent subtraction, thereby accelerating the reconstruction process. Moreover, it fundamentally relaxes the spectral constraints, shifting from strict separation of dual-wavelength spectra to the separability between difference-frequency terms and DC terms. The proposed algorithm requires merely half the carrier spatial frequency of the conventional Fourier transform algorithm, and enables high-quality reconstruction even under spectral overlap. Furthermore, the proposed algorithm not only simplifies the optical setup by removing components such as dichroic mirrors or polarization-based separation elements, but also maintains low computational complexity while achieving high-quality phase reconstruction in a single acquisition. Both numerical simulations and experimental results demonstrate the accuracy and effectiveness of the proposed approach.

Dual-wavelength off-axis digital holography enables direct unwrapped phase measurement on large phase samples, significantly reducing phase-unwrapping complexity. However, spectral crosstalk and noise amplification degrade reconstruction quality. Here, a dual-wavelength off-axis digital holography using frequency domain extension is proposed for high-quality phase-unwrapping measurements algorithm. A new frequency domain space with extended properties is generated through Fourier transform of Kronecker convolution between the dual-wavelength hologram and a 3 & times; 3 identity matrix. This operation simultaneously suppresses zeroorder term interference and crosstalk between the first-order term of both wavelengths, thereby eliminating phase jump in the results. Furthermore, the height information from the single wavelength, differential synthetic wavelength and additive synthetic wavelength is integrated within a least-squares optimization framework, significantly suppressing the noise in the reconstructed results. As a result, the high-quality reconstruction without phase jumps can be obtained by the proposed algorithm. Simulation and experimental results are conducted to validate the feasibility and effectiveness of the proposed algorithm.

Vision-based displacement measurement techniques have garnered significant attention in structural health monitoring due to their advantages, including non-contact operation and full-field-of-view capabilities. Among these techniques, correlation matching is particularly notable as it allows for sub-pixel displacement measurements from video data. However, extracting structural displacement from large volumes of video data (thousands of time-delayed frames per measurement) poses several challenges. These include high computational demands and potential errors, especially in the presence of sparse structural features (e.g., textures or contours) within the video. To overcome these challenges, this study introduces a feature-enhanced polyphase decomposition method for more accurate and efficient displacement measurement. By applying morphological dilation to video data contained structural displacement, a complex amplitude model is constructed, yielding multi-feature amplitudes with enhanced discriminability. Within a video -to-reality mapping framework, the decomposed impulses between these amplitudes are analyzed through the combination of phase spectrum correlation matching and polyphase decomposition. This yields a geometric motion model for decomposed impulses, resolves sampling parity ambiguity, and enables precise displacement recovery. Compared to existing spatial- and frequency-domain correlation matching techniques, the proposed method significantly reduces computational time while maintaining high accuracy. It also maintains robust performance under varied conditions, including video s with sparse features, limited visibility, or complex backgrounds. The method’s superiority has been demonstrated through extensive simulations and experiments. In challenging low-contrast complex scenarios, it achieved a correlation coefficient value of 95.54% and a root mean square error value of 3.69, significantly outperforming existing methods. Moreover, it processes data up to 8 times faster than traditional methods.

Digital holographic microscopy (DHM) is a powerful quantitative phase imaging method, enabling label-free 3D reconstruction with nanometer-scale axial resolution. However, the field of view (FOV) of off-axis DHM is limited due to the sampling limitations of sensors. To solve this problem, a DHM with an extended FOV is set up by introducing a diffraction grating in the object arm to generate three angularly separated object beams (-1st, 0th, +1st diffraction orders), each of which encodes complementary spatial information from distinct FOV regions and has different inclined angles. These beams are interferometrically multiplexed with a common reference wave on the sensor, allowing single-shot capture of spatial information for a threefold FOV extension. Phase reconstruction via Fourier transform, spectral filtering, and complex amplitude division can retrieve all three FOV simultaneously. The validity and feasibility of this DHM are verified by conducting experiments on 1951USAF resolution targets, microlens arrays, and bee wings. The experimental results show that this proposed DHM achieves threefold FOV extension without spatial resolution degradation, which demonstrates its potential for high-throughput applications and extensibility to other computational imaging modalities requiring wide-field quantitative phase analysis.

Heart sound sensors are essential tools for the early detection of cardiovascular and respiratory diseases. The core component of these sensors is the acoustic transducer, whose performance in capturing heart sound signals critically influences the accuracy of subsequent analysis and diagnosis. However, conventional acoustic transducers typically employ rigid diaphragms with poor mechanical compliance, limiting their ability to effectively couple with low-frequency, low-intensity heart sound signals. In nature, the spider can capture acoustically induced air particle motion using a slender orb-web across a broad frequency range at maximal physical coupling efficiency. Herein, inspired by this mechanism, we present a fiber-optic microphone based on a bioinspired spider-web-like structure (BSS) for heart sound detection. The BSS, with its high mechanical compliance, closely depicts the motion of acoustic particles and employs Fabry-Perot interference to convert its mechanical vibrations into optical signals, minimizing interference during signal readout. This sensor achieves ultra-high sensitivity (6714.29 nm center dot Pa-1 at 100 Hz), exceptional low-frequency response down to 1 Hz, and inherent directionality to sound, proving its suitability for the rapid detection of weak, low-frequency heart sound signals. This bioinspired sensing approach provides a novel and effective strategy for high-fidelity detection of weak, low-frequency acoustic signals.

We present a temperature and pressure dual-parametric spectral demodulation methods for Extrinsic Fabry-Perot Interferometer and Fiber Bragg Grating (EFPI-FBG) sensors based on deep learning, significantly reduced the influence of cross-talk between the parameters. An experiment was designed in this study to collect 36 sets of spectral data by regulating the parameter values with the pressure controller and the temperature chamber. Afterward, the dataset was properly classified and used to train the dual-branch feature extraction neural network, which utilized the Dense Convolution Network (Dense Net). The final results showed that the mean absolute errors (MAE) of temperature and pressure is 0.907 degrees C and 0.035 MPa. It takes only 0.38 ms to demodulate a spectrum. Moreover, the neural network model possesses generalization ability, providing reliable demodulation values for unknown spectra. Finally the study achieved real-time and accurate dual-parameter demodulation. Due to the ability of deep learning algorithms to extract information from raw spectra and automatically learn the complex features of the spectra, this method is expected to exhibit good demodulation performance even with more complex spectra in the future, allowing for more accurate parameter measurement.

Significance Vibration monitoring is a critical component ensuring the safe operation of engineering systems in aerospace,industrial manufacturing,and civil engineering.In these fields,abnormal vibrations can shorten the service life of precision aircraft components,degrade the quality and accuracy of mechanical processing,and threaten the health of structures such as bridges and buildings,potentially causing severe safety incidents.With ongoing advancements in engineering technology,the application scope of vibration monitoring has expanded to high-end equipment and intelligent systems including wind turbines,electric vehicles,and rail transit,playing an increasingly vital role in high-frequency dynamic response evaluation and structural fatigue warning. Traditional contact sensors face challenges such as loading effects and limited spatial resolution,while non-contact methods like laser vibrometers are constrained by spatiotemporal synchronization and cost in full-field measurements.In contrast,optical flow-based video vibration measurement offers advantages of non-contact operation,no pre-processing requirements,high spatial resolution,and strong real-time performance.By analyzing changes in image pixel intensity or phase to map structural displacement,it is less dependent on texture and well-suited for complex structures and high-frequency excitations,effectively addressing limitations of conventional methods.Therefore,the research and development of this technology not only provide innovative theoretical and technical support for engineering vibration monitoring but also hold significant practical importance and broad application potential in advancing dynamic evaluation of high-end equipment and structural health monitoring. Progress Considering the diverse demands for measurement speed and accuracy in engineering applications,various optical flow methods have been continuously refined and optimized,evolving toward a balance of efficiency and robustness.With sustained improvements in algorithm performance and expanding application domains,optical flow is increasingly recognized as a key technology with broad practical significance in vibration monitoring. In terms of method classification,intensity-based optical flow has seen significant performance improvements through enhanced signal processing and robustness mechanisms.For instance,Javh et al.integrated frequency-domain techniques to achieve near-10 kHz mode shape identification in metallic beams,successfully extracting eight modal parameters(Fig.7).Li et al.proposed an adaptive spatial filtering algorithm that selectively extracts valid pixels,enabling accurate modal separation under strong noise(Fig.8(a)).Chen et al.introduced a camera motion compensation method that effectively suppressed environmental disturbances during measurements(Fig.9(c)). Phase-based optical flow has achieved a dual breakthrough in both accuracy and computational efficiency.Miao et al.optimized the complex Gabor filter to reduce phase errors,while Ai et al.proposed an automatic scale selection strategy that enhances measurement precision by matching motion amplitude with object size(Fig.11).Additionally,Shan et al.'s temporal phase division method and Liu et al.'s differential phase optical flow(DPOF)significantly simplified the computational pipeline and improved processing speed(Fig.12). Deep learning-based optical flow further enhances full-field displacement estimation under weak textures and occlusions.Architectures such as SubFlowNet(Fig.15)and recurrent all-pairs field transforms(RAFT)(Fig.16),coupled with physics-consistent loss functions,demonstrate robust performance in complex scenarios.Research by Su et al.and Guo et al.has extended the applications of these methods to power equipment monitoring and challenging lighting conditions. Moreover,the integration of optical flow and video motion magnification has become a major innovation point.Shang et al.used motion magnification to identify multi-order vibration modes;Yang et al.enhanced frequency recognition through filter optimization;and Kang et al.realized full-field transverse vibration measurement of rotor systems via adaptive parameter tuning(Fig.18). In engineering applications,optical-flow-based video vibration measurement technology has been effectively validated in modal parameter identification,damage detection,and long-range monitoring tasks.For example,in modal tests of simply supported beams and air compressors,high-order modes were successfully extracted(Fig.19,Fig.20);in damage detection of acrylic frames and compressors,precise defect localization was achieved(Fig.21,Fig.22);and in long-range monitoring scenarios involving antenna towers and cable-stayed bridges up to 175 meters away,the method demonstrated high accuracy and efficiency in non-contact,full-field vibration measurement(Fig.23,Fig.24). Conclusions and Prospects Optical-flow-based video vibration measurement technology has emerged as a powerful technique for non-contact structural vibration measurement,achieving notable advances in accuracy,robustness,and engineering applicability through ongoing algorithmic improvements and integration with video magnification and deep learning.It enables high-resolution identification of complex mode shapes and micro-vibrations,and has demonstrated strong potential in modal testing,damage detection,and long-range monitoring. Despite its progress,challenges remain in handling large displacements,achieving accurate full-field calibration,reconstructing 3D vibration fields from monocular input,and maintaining robustness under complex environmental conditions.Future research should focus on developing efficient large-displacement matching algorithms,adaptive scale calibration models,and multi-view reconstruction strategies.Integrating the method with edge computing and intelligent systems will be key to advancing from lab-scale validation to real-world deployment,offering reliable tools for structural health monitoring and intelligent equipment maintenance.

Visual vibration measurement is an emerging non-contact technique capable of high spatial resolution and full field. Phase-based motion estimation (PME) has gained widespread attention in visual vibration measurement due to its robustness and insensitivity to measurement environment. However, most existing PME methods, including both filter-based and derivative-based approaches, rely on empirically tuned parameters for either filter design or analytic signal construction, which limits their accuracy and generalizability. To address this, this paper proposes a self-determining multiderivative-enhanced phase motion estimation (MDPME) framework. By leveraging structured multi-order spatial derivatives of image intensity, MDPME constructs a parameter-free complex analytic signal and intrinsically determines the phase-to-displacement scale factor, eliminating the need for manual tuning or empirical calibration. This formulation transforms PME from a parameter-dependent approach into a self-determining framework while preserving the computational efficiency of derivative-based methods. Experimental results demonstrate that MDPME achieves a correlation coefficient exceeding 99.13% with reference sensor data, underscoring its superior performance over established PME techniques.

Out-of-plane vibration modal analysis is essential in structural health monitoring (SHM) for identifying dynamic anomalies associated with structural damage. However, vision-based approaches using depth cameras often suffer from missing depth data and noise interference, limiting their applicability in modal identification. To address these challenges specifically for slender one-dimensional (1D) beam-like structures, this study proposes a noise-robust depth reconstruction-based modal identification method (DRMI). The method integrates directional depth completion to restore missing regions, adaptive moving average filtering for noise suppression and computational efficiency, and least-squares polynomial fitting to mitigate quantization noise induced by limited depth resolution. Differential amplification is further applied to enhance high-order modal components, followed by Hankel dynamic modal decomposition for modal parameter extraction. Validated through both numerical simulations and laboratory experiments on slender beam-type structures, the proposed DRMI method enables accurate and noise-resilient identification of high-order and multi-band vibration modes directly from depth sequences. Compared with the state-of-the-art pyramid reconstruction method, DRMI improves the average modal assurance criterion by 8.6% and reduces the root mean square error by 64.8%. These results demonstrate that DRMI provides a physically consistent and noise-robust framework for out-of-plane modal analysis of slender 1D structures in practical SHM scenarios.

Sleeve bearings are critical components for ensuring the safe operation of ships, yet they are prone to wear under harsh working conditions, thereby affecting the propulsion performance of vessels. Precise monitoring of bearing wear remains a challenge, particularly for early-stage detection where high-accuracy techniques are lacking. This article proposes a high-accuracy, direct measurement method for sleeve bearing wear utilizing a chirped fiber Bragg grating Fabry-Perot (CFBG-FP) sensor. Leveraging the linear relationship between Fabry-Perot fringe count and grating length, we developed a dedicated wear demodulation algorithm based on the fringe-counting approach, ensuring accurate interrogation by employing an automated multipeak detection technique. Experimental results demonstrate that the sensor achieves a measurement error of 0.2 mm, a resolution of 0.04 mm, and a wear measurement range of 11 mm. This work closes the precision gap in early-stage sleeve bearing wear monitoring with a novel sensor system, providing a robust solution for maritime applications.

This paper proposes a fiber-optic Fabry-Perot (F-P) strain sensor utilizing a rhombic hinge mechanism, successfully achieving high sensitivity detection of aluminum alloy specimens at the nano-strain (nE) level on standard samples. The diamond-shaped hinge amplification mechanism centrally constrains the input strain in the sensitive region of the F-P cavity for linear amplification, effectively addressing the sensitivity degradation issue caused by strain dispersion in standard F-P and FBG sensors. Theoretical research indicates that the mechanical amplification structure proposed in this study can enhance the amplification ratio while preserving a low input stiffness, hence enabling nE-level detection. The integration of the F-P interferometric demodulation approach enhances range while preserving detection accuracy. Static experiments indicate that the system may attain strain measurements ranging from 1.43 n epsilon to 14.29 mu epsilon (1.43 x 10-9 to 1.429 x 10-5). The low-frequency dynamic test indicated that the strain response interval of the fabricated sensor was 1.43 n epsilon to 1428.57 n epsilon within the frequency range of 0.5-2 Hz. The dynamic test sensitivity for both static and low-frequency testing was 1.94 nm/n epsilon, aligning with the theoretical sensitivity. The least observed strain of the fabricated F-P strain sensor is 0.142 n epsilon. The sensor possesses significant potential applications in aerospace and composite structure health assessment.

Spider webs can couple acoustic signals with maximum physical efficiency over a broad frequency range. Inspired by the acoustic flow sensing mechanism of spider webs, a fiber-optic Fabry-P & eacute;rot (F-P) hydrophone is proposed. The sensitive diaphragm of the hydrophone adopts a bionic web-like structure (BWS), where the hub diaphragm of the structure and the fiber end-face form an F-P cavity. Finite element simulation is employed to optimize the geometric structure of the BWS and a full-phase spectrum demodulation method is used to demodulate the interference spectra. The test results demonstrate that the BWS hydrophone features a bandwidth of 10-2000 Hz, a sensitivity of 52.4 nm/Pa at 100 Hz and a minimum detectable pressure of 0.44 mPa/Hz(1/2), and it also exhibits favorable acoustic directivity. Furthermore, due to the hollowed-out structure of the BWS, the hydrophone is immune to hydrostatic pressure. This fiber-optic hydrophone based on the acoustic flow mechanism provides a novel strategy and solution for low-frequency, high-sensitivity underwater acoustic signal detection.

This study designs and fabricates a differential fiber-optic Fabry-P & eacute;rot interferometric (FPI) acoustic sensor for detecting small rotor UAVs. By placing two optical fibers with a 45 degrees inclination parallel on both sides of a polyethylene terephthalate (PET) diaphragm surface, a differential Fabry-P & eacute;rot (F-P) cavity structure is formed. Results indicate that this differential configuration enhances the signal-to-noise ratio (SNR). We employ Mel-frequency cepstral coefficients (MFCCs) as acoustic features and construct a convolutional neural network (CNN)-long short-term memory (LSTM) model for binary classification in UAV detection. Signals from a DJI Mini3 UAV at a distance of 200 m are collected using the fabricated sensor for recognition. The results demonstrate that after adopting the differential structure, the detection accuracy for UAVs reaches 99.24%, surpassing the 96.97% achieved without the differential structure, thereby validating the effectiveness of the differential design in enhancing the detection capability for weak UAV signals. This study provides a novel and effective solution for long-distance acoustic monitoring of UAV signals.

To meet the requirements for high-sensitivity and high-stability acceleration measurement in space-constrained and harsh environments, this article proposes and demonstrates a miniaturized Fabry-Perot (F-P) accelerometer based on a 45 degrees differential fiber interferometric configuration. A microscale elastic sensing structure with dimensions of 4.5 x 4.5 x 2 mm was designed and fabricated using beryllium bronze combined with chemical etching. Two symmetrically arranged 45 degrees fibers and the diaphragm form a differential Fabry-Perot interferometric (DFPI) cavity, which enables high-precision differential detection of cavity length variations. This configuration effectively enhances the signal amplitude and improves the system signal-to-noise ratio (SNR). The mechanical characteristics and frequency response of the sensor were optimized through analytical modeling and finite element simulations. An experimental platform was established to systematically evaluate the sensor performance. Experimental results demonstrate that the dual-fiber differential structure provides significantly better performance than the single-fiber configuration. The sensor exhibits a resonant frequency of 3100 Hz and an effective operating bandwidth of 0-1200 Hz. The average sensitivity within the working band reaches 60.46 nm/g. The acceleration resolution is 0.00307 g. The SNR is improved by approximately 3 dB compared with a single-channel structure. The experimental results confirm the accuracy of the theoretical design and finite element analysis. The proposed accelerometer achieves a favorable balance among miniaturization, high sensitivity, high accuracy, and high stability. It provides a feasible solution for precision vibration monitoring and structural health monitoring in space-limited environments.

In-line digital holography enables full image sensor bandwidth reconstruction, but its quantitative imaging capability is hampered by twin image artifacts resulting from spectral aliasing. Here, an in-line digital holography method based on dual-wavelength coupled phase retrieval (DwCPR) is proposed for non-crosstalk and high-resolution imaging. In this method, a single image sensor is used to capture an off-axis hologram and a composite in-line hologram at the same position through two exposures. The phase reconstructed from the off-axis hologram serves as the initial estimate, and an iterative model is established by coupling the dual-wavelength composite in-line hologram. The high-resolution reconstruction is achieved through the application of dual-wavelength plane amplitude constraints and energy conservation principles. The proposed method can avoid pixel errors caused by multiple sensors and minimize the required measurements across multiple planes. More importantly, the combination of energy conservation and multiple amplitude constraints significantly enhances the reconstruction quality.

Video-based vibration modal identification is an emerging technique for structural dynamics characteristics measurement. However, identifying physical modes typically requires extensive manual spectral interpretation by experts, or the over-estimation of candidate modes. As a result, automatic and efficient modal determination remains challenging when dealing with mode-mixed video data that contains complex mode components and large amounts of information. Here, a novel approach is presented by introducing binary decomposition tree (BDT) to adaptively separate and identify vibration modes. In the BDT framework, the original mode-mixing signal is treated as the root node, procedure modes are represented as branch nodes, and physical modes are depicted as leaf nodes. Each binary decomposition forms a subtree, and the decomposition process progresses from the mode-mixing root to the single-mode leaves. This method allows for the extraction of physical modes without manual intervention or the over-specification of dynamic behavior. A series of simulations and experimental tests were conducted to evaluate the automaticity and effectiveness of the proposed method. The results indicate that the proposed method achieves a correlation coefficient of >= 99.95 % and a modal assurance criterion of >= 99.97 % in identifying four mixed modes, with a processing speed about ten times faster than current methods. These metrics demonstrate that the proposed method surpasses existing approaches in both identification accuracy and computational efficiency.

Aiming at the problems that traditional ship underwater acoustic target recognition methods are difficult to achieve high recognition accuracy and have low robustness in complex underwater environments, this paper proposes a target recognition model that integrates multi-scale residual convolution and dual attention mechanism(MSR-DAN). A multi-scale residual convolution module is designed to capture the local and global features of acoustic signals. Combined with channel attention and spatial attention, the adaptive allocation of feature weights is achieved to enhance the expression of the characteristic information of ship radiated noise. The experimental results show that this method can achieve a recognition accuracy rate of 97.86% on the public ship dataset. Under the condition of low signal-to-noise ratio, it can also obtain an accuracy rate superior to other models.