With the continuous advancement of communication systems towards high-speed and ultra-wideband, there has been a growing research focus on designing high-performance Raman amplifiers tailored to this progress. However, designing high-performance Raman amplifiers is challenging due to the complex non-linear relationship between output Raman gain, noise, and pump parameters. Traditional numerical optimization methods are not efficient in solving this problem. To address this, this paper proposes a second-order Raman fiber amplifier (RFA) gain and noise prediction model using convolutional neural network (CNN) and long-short term memory (LSTM) . The impact of different prediction model performances on the design of Raman fiber amplifiers is investigated, and to optimize the model using the sea horse optimizer (SHO) algorithm to accurately reflect the mapping relationship between pump parameters, fiber length, and the target gain and noise distribution. Experimental results show that the proposed model has a root mean square error of only 0.043 1 and 0.022 4 dB in gain and noise prediction, with an error between the predicted and target values below 0.25 dB and an average consumption time of less than 0.133 7 s. This design provides methods and ideas for the rapid design of Raman fiber amplifiers in the future.

Monitoring refractive index and axial strain is crucial in engineering applications. To achieve a large range sensor for the measurement of dual parameters of refractive index and strain, we propose a fiber sensor based on a tapered thin-core fiber (TCF). A compact all-fiber Mach-Zehnder interferometer (MZI) is fabricated by fusing a 2 cm length of TCF between two standard single-mode fibers (SMFs), charging and tapering it in the center. In our experiment, we extend the measurement range of the refractive index by using propylene glycol solution instead of sodium chloride solution. The results show that the highest sensitivity of the refractive index is -20.95 nm/RIU in the range of 1.351 1—1.380 3 and can reach -294.32 nm/RIU in the range of 1.384 3—1.446 6. In the strain range of 0—272 7 , it exhibits a sensitivity of -0.6 pm/, indicating that the structure maintains strong robustness even with a large refractive index range and high sensitivity. The device is simple to manufacture, compact, and has application value in refractive index and strain measurement.

Camouflaged object detection (COD) aims to detect camouflaged objects hidden in complex backgrounds. Due to the characteristics of camouflaged object, such as similar foreground and background textures and low-contrast edges, existing methods often produce blurry edge of prediction image and miss small object regions. Therefore, this paper proposes an edge information guided network (EIGNet). First, the edge of the camouflaged object is explicitly modeled through low-level and high-level features, which fully extract the edge features of the object to guide subsequent feature representations. Then, a dual-branch structure is used to process different dimensions of camouflaged object. The global branch is used to extract global contextual information to emphasize the global contribution of large objects, while the local branch is used to mine rich local low-level clues to enhance the feature representation of small objects. Finally, a top-down manner is used to gradually aggregate adjacent layer features to obtain a prediction image with fine edges and complete regions. Experimental results on three camouflaged datasets show that our method outperforms 15 other models, with a mean absolute error (MAE) of 0.044 on the NC4K dataset.

For the tracking problem of moving pedestrians in indoor environment, this paper proposes a moving target tracking method based on the fusion of monocular vision and depth vision sensing information. Firstly, we use monocular vision image information to achieve target detection and complete semantic segmentation of the target. Then, we jointly calibrate monocular vision with depth vision to obtain the depth information of the recognition target by the depth camera. Finally, a moving target tracking method based on multi-camera information fusion is designed to address the problems of large error in monocular vision perspective-3-point (P3P) solution when the target is far away and slow frame rate of depth camera recognition, and the distance information obtained from monocular vision and depth information extracted from depth vision are asynchronously fused to achieve accurate estimation of the target motion state. The results show that the algorithm achieves the effect of real-time detection and tracking of pedestrian target positions within the field of view, with efficient target recognition and tracking capabilities.

Aiming at the problems of poor overall contrast, unclear contour details and poor visual effect caused by weather in remote sensing images, a scheme which is based on non-subsampled contourlet transform (NSCT) algorithm and combined with Retinex and bilateral filtering, is proposed to achieve image dehazing enhancement. Firstly, NSCT decomposition is performed on each channel of the hazy image to obtain high frequency and low frequency subbands. Then, Retinex algorithm and linear enhancement processing are used in the low frequency subband to improve the image brightness uniformity and gray dynamic range. The bilateral filtering and gradient enhancement are used in the high frequency subbands to filter out noise while retaining and enhancing the image edge contour. Finally, the low frequency and high frequency subbands are fused to achieve dehazing enhancement, and the white balance algorithm is used to maintain the consistency and accuracy of the image color. The experiments of different algorithm show that the average gradient is improved by 17.51, and the mean value is moderate. In the process of denoising, the proposed algorithm can effectively enhance high frequency details and improve contrast, leading to noticeable improvement in image quality.

Segmentation plays a crucial role in the computer-aided diagnosis of keratoconus. This paper proposes an accurate segmentation algorithm for the corneal deformation area in video images of corneal force deformation, based on a fully convolutional architecture integrated with an attention mechanism (AM). It includes three key technologies: skip connections (SC), residual convolution (RC), and a fully convolutional architecture integrated with global AM. Skip connections effectively enhance the model's ability to learn complex contour details, while RC allows for the construction of deeper feature models while retaining basic image features. The global AM improves the segmentation accuracy of the model by extracting refined feature maps from each convolutional and deconvolutional block. By enhancing and highlighting key areas, it has been demonstrated that more accurate segmentation of the corneal deformation area effectively improves the early diagnosis accuracy of keratoconus.

Tensor principal component analysis (TPCA), as a data dimensionality reduction algorithm aimed at representing high-dimensional tensor data in low dimensional subspaces, has been widely applied in multiple machine learning fields. However, the L1-norm loses the rotation invariance and the existing TPCA algorithms adopt a single optimization objective, which only optimizes the projection distances and ignores the optimization of the error tensors. Thus, even though these algorithms have a certain degree of robustness, they still perform weakly. To address these issues, this paper proposes a ratio model for dual-objective optimization. This model is inspired by the formula for the area of a right-angled triangle, which optimizes the height on the hypotenuse to achieve dual-objective optimization with maximum projection distances and minimum reconstruction errors, called the area projection model. Then, based on the projection model, this article adopts a preprocessing technique of blocking recombination and proposes a block tensor PCA with F-norm based on area projection (area-BTPCA-F) algorithm. This algorithm not only preserves rotation invariance, but also fully considers error tensors. In response to noise information, blocking recombination technique has greatly improved the robustness of the algorithm. Finally, experiments on six color datasets with different noise validate the proposed algorithm, showing improvements in average reconstruction error (ARCE) and classification rate. The algorithm demonstrates strong robustness compared to other existing TPCA algorithms.

In this study, the radiation characteristics of nonlinear Thomson scattering were investigated using the Lagrange-Gaussian model. Through analysis and experiments, we obtained the spatial motion trajectories of individual electrons and generated three-dimensional radiation data within the laser field. The results indicate that, under the conditions of a fixed operating wavelength, waist radius, and pulse width, electrons within a tightly focused linearly polarized laser field exhibit oscillatory motion in the x-z plane, forming a "zigzag" pattern, with the maximum displacement along the x-axis. Additionally, for the first time, we employed a time-resolved observation method to observe the three-dimensional radiation changes resulting from the interaction between the laser and electrons. We found that the normalized radiation energy exhibits different growth rates as the normalized time increases. Furthermore, within the laser field, the angles and of the normalized peak energy are not fixed; only when the normalized peak energy tends towards its maximum value of 1.45 hundred million, the observation angles become fixed. These results are of significant importance for understanding nonlinear Thomson scattering and advancing research on X/-ray generation in optical laboratories.

In order to improve the adaptability of laser fuse in smoky environment, the echoing characteristics of linear frequency modulated (LFM) pulse laser fuse is studied. The laser scattering model in smoke is established by optimizing the calculation speed of Monte Carlo algorithm. The results show that the echoing signal of LFM pulse in smoke is composed of three types. They are low frequency (LF), high frequency (HF) and target signal. When the smoke concentration is 1.4 mg/m3, the target signal is completely attenuated. Although the LF of smoke has little effect on target signal detection, when the smoke concentration is greater than 1.2 mg/m3, the aliasing of smoke HF and target signal spectrum causes the laser fuse to generate false alarm signals. Compared with traditional pulse lasers, LFM pulse lasers can adapt to higher concentrations smoke. The wider the bandwidth, the stronger the ability to resist smoke interference. The research results provide important support for the design and signal processing of LFM pulse laser fuses.

In order to manipulate terahertz wave flexibly, a terahertz tunable coding metasurface unit cell structure is designed based on the Pancharatnam-Berry phase principle and the use of graphene material. It consists of a double-open -shaped metal structure on the top layer, a polyimide substrate embedded in a double-layer graphene structure and a metal layer, and its reflection amplitude can be dynamically tuned by modulating the Fermi energy of the bilayer graphene structure. Based on its amplitude tunable characteristic, a coding sequence with rotational phase gradient characteristic is designed to generate an amplitude tunable vortex beam. Combined with the Fourier convolution theory, a coding metasurface capable of emitting a double vortex beam is designed ulteriorly, and the maneuverability of vortex beam numbers and scattering angles is greatly improved in this way. The device designed has potential applications in terahertz high-speed communication system.

So far, YAG: Ce3+ yellow phosphor excited by blue light still holds a large share in the white light emitting diode (LED) market. but it has inherent defects of red light components' deficiency, so the obtained white light has a low color-rendering index and a high color temperature. However, the white light that relies on rare earth ions to achieve the harmony of red, green, and blue primary colors can overcome this drawback effectively. This paper applied traditional hydrothermal reactions to synthesize red phosphor, a series of Sm3+ ion doped BaZrO3 in the same phase as the substrate in one step and whose test results of X- ray diffraction (XRD), scanning electron microscopy (SEM), particle size distribution, Fourier transform infrared spectrum (FT-IR) and fluorescence spectrometer (FL) showed that the obtained products all exhibit a cubic perovskite structure and furthermore, after more heterovalent Sm3+ replaced Zr4+ ion, parameters, such as grain size, crystal cell parameters, and the lattice strain, were increased. In addition, as the concentration of Sm3+ ions increased, the infrared spectrum peak gradually shifted blue until reaching its maximum value at Sm3+=3.0 mol%, which was consistent with the result calculated by molecular vibrational frequency. Finally, the relationship based on the relative intensity between the broad emission band in the spectrum and the characteristic luminescence peak of Sm3+ verified that luminescence from defects existed in BaZrO3 crystal, and after Sm3+ ions were introduced and the energy transmission between BaZrO3 and Sm3+ could be realized.

Metal artifact leads to reduced quality of computed tomography (CT) images, which can severely degrade segmentation accuracy. To address this problem, a segmentation network is proposed to resist metal artifact interference. This network uses composite connected dual-stream encoder structure with two backbones for feature extraction from disturbed and undisturbed CT images, respectively. The composite connection structure integrates the features extracted by the encoders on the two backbones. A Transformer-based focal self-attention (SA) mechanism block is developed to encode global multi-scale information. The training process of the network is optimized using hybrid loss and ancillary supervision. The experimental results show that this network on metal artifact data could reach 86.40%, 93.11% and 90.76% in average Dice coefficient, MIoU and Recall, respectively. The network has great anti-metal artifact interference effect in semantic segmentation for CT images, and achieves high segmentation accuracy without artifact reduction.

Fiber optic surface plasmon resonance (SPR) sensor has the advantages of high detection sensitivity, fast response speed and real-time online monitoring, and has a good application prospect in biology, chemistry, medicine and other fields. However, the traditional single-channel fiber SPR sensor has only one sensing area and can only perform one single-point measurement, which greatly limits its further development and application. The multi-channel fiber SPR sensor can make up for the shortcomings of traditional single-channel fiber SPR sensor because of its large detection flux, many detection substances, realization of temperature compensation, and elimination of various non-specific responses. The research of multi-channel fiber SPR sensor is of great significance to improve the detection sensitivity and detection accuracy of traditional single-channel fiber SPR sensor and expand its application fields. First of all, the principle of fiber SPR sensor is introduced. Then, the research progress of multi-channel fiber SPR sensor in recent years is presented. At the end, the future research direction of multi-channel fiber SPR sensor is prospected. Multi-channel fiber SPR sensor with new structure, high performance and multi-application is an important development direction of multi-channel fiber SPR sensor in the future, and the integration of new materials and new processes will surely bring the development and application of multi-channel fiber SPR sensor to a new level.