Optical Review最新文献

Infrared sensors are widely used in human action recognition because of their low light influence and excellent privacy protection. However, the traditional deep learning networks and training or testing methods tend to fall into the trap of local optimum because of the similarity between infrared image classes and the lack of discriminative features such as texture and depth, and thus obtain poor recognition results. To address this issue, we propose a novel human action recognition method based on similarity evaluation. This method innovatively transforms the traditional training and testing (verification) mode. First, we use a feature-to-feature training method to make the network pay more attention to the behavioral information that distinguishes the classes. Second, we design a Integrate Channel Attention Module(ICA) to enable Siamese network to focus on the areas of interest. Finally, we propose the Multimodal Similarity Evaluation Module (MSE). The module aims to address the fuzzy matching problem of feature areas. The contrast experiment results show that our method outperforms existing mainstream methods on several benchmark datasets. The excellent accuracy provides an innovative method for addressing various problems related to high similarity between classes.

Turbulence effect is an important factor leading to performance degradation of underwater wireless optical communication (UWOC) systems. In order to improve the reliability of the communication system, this paper adopts a UWOC system based on low-density parity-check (LDPC) code and pulse position modulation (PPM) to overcome the effects caused by turbulence, and proposes an improved Offset Min-Sum algorithm to combat turbulent channel interference. The generalized gamma distribution (GGD) distribution is used as the channel model, and the relationship between BER, SNR, different of LDPC decoding algorithms, and PPM order is quantified by simulation for different turbulence intensities.The results show that the performance of the improved OMS algorithm is optimal under weak, moderate, and strong turbulence; at ({sigma }_{I}^{2})=2.0399 and BER = 10^–6, the improved OMS algorithm with 4-PPM, 8-PPM, and 16-PPM modulation has a coding gain of 0.16 dB, 1.2 dB, and 4 dB, respectively, compared with the OMS algorithm.

In this paper, we report the theory of a self-coupling laser sensor array system for improving the signal-frequency inversion problem and the results of velocity measurements with this system. A self-coupling laser sensor is an interferometer that uses optical beats produced by the interference between the light returned from the laser target and the light in the active layer of the laser diode. Using wavelength modulation, this system can simultaneously measure multiple metrological quantities, such as the absolute distance to a target and velocity of a target. However, in a self-coupling laser sensor using wavelength modulation, the signal frequency is inverted and becomes negative if the Doppler shift of the returned light owing to the movement of the target is larger than the signal frequency when the target is stopped. In this case, it is impossible to detect the positive or negative value of the signal directly, resulting in a large measurement error. This has been regarded as a problem that limits the measurement dynamic range of the modulated self-coupling laser sensors. In this study, we propose a system to accurately detect the signal-frequency inversion and improve the measurement dynamic range. The proposed system detects the positive or negative value of the signal frequency from the relationship between the velocity and signal frequency obtained by irradiating multiple beams with different modulation frequencies and then recalculates the accurate measurement value. The measurement results reveal that this system can accurately measure the moving velocity of a target, even when the signal frequency is inverted.

In this study, we present a through-focus re-radiation simulation aimed at detecting scattering from semiconductor structures. We employ the beam synthesis propagation (BSP) module within the finite-difference time-domain (FDTD) method, optimizing the simulation of optical systems by reducing time and computational resources typically required for imaging and illumination. To validate the approach, we simulated scattering from Silicon nitride (Si₃N₄) lines on a silicon (Si) substrate with various defect sizes and types at a 193 nm wavelength. The results demonstrated the detection of specific defect signals and identified the limitations of detectable defect sizes. These findings are intended to serve as pre-processing data for predicting outcomes in through-focus scanning optical microscopy (TSOM) imaging.

Injection-molded lenses have an inhomogeneous stress-induced birefringence that can degrade optical performance. This paper presents a new approach for measuring and analyzing inhomogeneous anisotropic samples. The birefringence distribution is characterized by 3D index ellipsoids, and a tomographic reconstruction of this 3D distribution is developed from a linear line projection relationship between the spatially varying index ellipsoids and tomographic polarimetry. This forward representation enables a tensor-valued backprojection for reconstructing the birefringence distribution of an inhomogeneous anisotropic sample. In this approach, each index ellipsoid is represented by a Hermitian matrix, and the 3D birefringence distribution is defined as the distribution of these matrices. This paper is centered on the introduction of the fundamental algorithm and the presentation of a general solution by applying the Radon transform and the backprojection to a tensor field, without requiring specific parameters such as stress fields. Consequently, the computational approach presented in this paper demonstrates that, using 60 tomographic views, reconstruction errors for parameters that characterize spatially varying index ellipsoids remain less than 5%. Here, the error is defined as the ratio of reconstruction variation to the respective maximum values of the original distributions.

The demand for fast, accurate, and cost-effective methods for three-dimensional shape and color measurements has been increasing. Ideally, both the shape and color of an object should be obtained in a single shot. Color fringe projection profilometry allows single-shot 3D shape measurement; however, it faces challenges when applied to colored objects. The fringe patterns are attenuated, leading to inaccuracies in shape measurement, and the fringes obscure the object's color information. This study proposes a novel approach to address these challenges by using a deep learning-based ResUNet model. Our method uses two independently trained ResUNets to correct fringe distortions for improved shape measurement accuracy and to remove fringe patterns for color information extraction from the same captured images. The simulation and experimental results demonstrate the effectiveness and applicability of this approach for single-shot 3D shape and color measurements.

In various sports, the motion information of athletes is often measured for monitoring and evaluation, and the direct use of common optical equipment to determine the position and orientation of moving objects according to geometric information in a scene has become an important research topic in image understanding. As many sports grounds have a center circle and halfway line, we propose an algorithm that first obtains constraints on the image of the circle center by using homography based on the geometric properties of the circle perimeter and corresponding circumferential angle. Then, the vanishing line is obtained from the image of the circle center and the complete circle image based on the pole-polar relation with respect to the camera internal parameters. By decomposing the circle image, the camera external parameters are obtained to determine the homography matrix from a spatial point to an image point. The camera at the edge of the moving field is calibrated according to the duality of the conic and homography matrices. Using the homography matrix between a point on the moving ground plane and the corresponding image point, the coordinates of the measured point can be recovered to estimate the pose (i.e., position and orientation) of a moving target.

This study investigates the entanglement and squeezing characteristics of light generated by a non-degenerate, coherently driven three-level laser in an open cavity, coupled to a two-mode vacuum reservoir through a single-port mirror. By normal ordering the noise operators, we simplified the calculations and derived the evolution equations for the atomic operators using the master equation. From the steady-state solutions, we determined the average of the photon number, the quadrature variance of radiation, entanglement, the normalized second-order correlation of the cavity radiation, the linear correlation coefficient between the two modes, and fluctuations in intensity difference. Our findings indicate that higher spontaneous emission rates significantly decrease the average photon number, while the amplitude of the pumping mode interacting with the parametric amplifier ((varepsilon)) increases it. Enhanced squeezing is observed with increasing ((varepsilon)), reaching a peak at (varepsilon = 0.03) ((72.6%)). Moreover, spontaneous emission enhances squeezing. A direct correlation between squeezing and entanglement is found, with greater squeezing associated with increased entanglement. These insights have significant implications for advancing quantum technologies, such as quantum communication, where controlled squeezing and entanglement improve secure communication channels and signal-to-noise ratios, quantum computing, where they enhance error correction protocols and gate operation efficiencies, and quantum sensing, where they increase sensitivity for more precise measurements of physical quantities.

A novel self-powered ultraviolet (UV) photodetector (PD) based on a CuI/MgZnO heterojunction modified by an ultrathin Cu₂O layer has been fabricated by the successive ionic layer adsorption and reaction (SILAR) method. Compared with the CuI/MgZnO self-powered PD, the optimised heterojunction PD (CuI/Cu₂O/MgZnO) exhibits significantly improved self-powered properties. Under 325 nm UV light at an intensity of 450 µW/cm², the CuI/Cu₂O/MgZnO heterojunction PD shows exceptional photoelectric performance, featuring a high photo-to-dark current ratio of 1611, a large responsivity of 48.43 mA/W, and rapid rise and decay times of 261 ms and 890 ms, respectively, without any external power supply. Incorporating the Cu₂O interface layer results in notable enhancements in responsivity and detectivity compared to the heterojunction without the Cu₂O layer. This improvement is attributed to heterojunction interface contact, energy band engineering, and the tunneling effect. The Cu₂O layer expands the depletion zone and promotes charge separation. Due to its thinness, charges can tunnel through the Cu₂O layer from one metal electrode to another. Furthermore, the interfacial Cu₂O layer can alter the valence band offset and the conduction band offset of the p-CuI/n-MgZnO junction, enhancing carrier transport between MgZnO and CuI. These results lay the groundwork for using self-powered MgZnO-based heterojunction photodetectors in light-based devices in the future. They also demonstrate the potential of designing novel heterojunctions to create high-performance self-powered PDs for UV detection.

In the industrial production of steel materials, various complex defects may appear on the steel surface owing to the influence of environmental and other ambient factors. These defects are often accompanied by large amounts of background texture information. Especially, some defects with the low resolution and small size are prone to false alarms and missing detections. Aiming to address the issues of these specific defects, this paper proposes a bidirectional cross-scale feature fusion network combined with non-stridden convolution for steel surface defect detection. First, to improve the model’s inference speed and reduce the number of parameters, a simple yet effective convolution (PConv), the core component of FasterNet, is introduced in the feature extraction module instead of the traditional ResNet operator. Second, the bidirectional crossing (BiC) module is embedded to construct a bidirectional cross-scale feature fusion network (BiCCFM), which provides more accurate localization clues to enhance the feature representation on small targets. Finally, combined with non-stridden convolution, the SPD-Conv module is developed to aggregate the detection performance of small targets in low-resolution images. Comprehensive experimental results on the public NEU-DET dataset validate the effectiveness of the embedded modules and the proposed model. Compared with other state-of-the-art methods, the proposed model achieves the best accuracy (74.2% mAP @ 0.5) while maintaining a relatively small number of parameters.