Based on the research on the structures of two types of photonic crystal functional devices,optical beam splitters,and filters,this paper designs a novel two-dimensional triangular lattice photonic crystal filtering beam splitter,which can simultaneously achieve power beam splitting and filtering functions.The finite difference time domain method are used to analyze the transmission characteristics of light waves in the new structure.The results show that the proposed structure successfully integrates the characteristics of filtering and beam splitting,and can simultaneously achieve good filtering and beam splitting functions under different incident wavelengths.For a wavelength of 1.527 μm and 1.550 μm,the incident light wave is basically symmetrically split and output,with transmittance of 66.77% and 95.22%,respectively.Meanwhile,for the incident light waves of these two wavelengths,the structure also has good filtering performance,with corresponding filtering efficiencies of 70.62% and 92.76%,respectively.In addition,for a wavelength of 1.577 μm,the structure mainly functions as a filter,with a filtering efficiency of 98.11%.

In this paper,based on Fabry-Perot cavity interference theory and elastic diaphragm resonant system,a new fiber optic vibration sensing system is designed.Its vibration sensing principle is analyzed theoretically,and the finite element model is constructed by Solidworks.The modal analysis and vibration characteristic response analysis of the sensor are carried out.The multi-peak demodulation algorithm is used to realize the real-time acquisition of the absolute cavity length,which has the advantages of high resolution and large dynamic range,and can adapt to the situation that the cavity length varies greatly.The dynamic response characteristics under the periodic frequency change are investigated through experiments,and the experimental results show that:the effective frequency range of the sensor is 5—200 Hz,the resonant frequency is 260 Hz,the low frequency band of 5—20 Hz has a flat amplitude-frequency response curve,the lowest detection acceleration is 0.1 m/s2,the sensor works stably under 0.1—10 m/s2 acceleration,the axial sensitivity is 8.19 μm/g,and the acceleration resolution is 1.5×10-5 g.It can be used to monitor the low frequency infrasonic wave of pipeline leakage.

In order to avoid the large errors in calculating the fiber Bragg grating (FBG) center wavelength caused by the unstable modulation depth when the phase generated carrier (PGC) demodulation method is used to demodulate the signals of FBG vibration sensors,a PGC-arctangent-peak finding (PGC-Arctan-PF) joint demodulation algorithm is proposed.Firstly,the output signal of the optical path multiplied with the fundamental frequency and the two-fold frequency of the carrier signal respectively.Then passing them through the low-pass filter (LPF) to obtain quadrature signals of different amplitude containing the vibration information to be measured.After that,the peaks of this orthogonal signal are searched one by one.The average amplitude of the two signals is calculated separately.The average amplitude is cross multiplied of the two signals to make the quadrature signals equal in amplitude.Finally,the frequency and amplitude of the stretched FBG are accurately solved after an arctangent operation and high-pass filter (HPF) processing.The feasibility of the PGC-arctan-PF algorithm is verified by simulation and experiment.This algorithm can accurately solve the vibration signal from 100 Hz to 1 000 Hz,and the relative error of FBG wavelength calculation is less than 0.81%.The PGC-Arctan-PF algorithm can effectively avoid the influence of modulation depth fluctuation on the solution accuracy.

In order to verify the relationship between spectrum and lettuce water loss rate,chlorophyll and freshness,this research employed Vis-NIR (280—1 100 nm) hyperspectral imaging system to collect the spectral data of lettuce under different storage conditions with different storage times,and measured the water loss change and chlorophyll content at the same time.The original spectra were preprocessed by standard normal variate (SNV),multicative scatter correction (MSC),Savitzky-Golay (S-G),which were further analyzed by the first-order differential and second-order differential methods and then the characteristic wavelength was selected by competitive adaptive reweighted sampling (CARS) algorithm and successive projection algorithm (SPA).The prediction models were established with the original spectrum and characteristic wavelength respectively by partial least squares support vector machine (PLS-SVM) and BP neural network (BPNN) to predict the freshness and water loss rate of lettuce.The results show that BPNN has a satisfactory predicting accuracy,and the prediction accuracy of water loss rate reaches 82.5% and the prediction accuracy of freshness is as high as 95%.Finally,spectral changes at different storage times were analyzed by principal component analysis (PCA),and the principal component images were extracted to visualize the wilting process,then,it demonstrated that fresh lettuce under soaking and refrigerated storage conditions showed a significant delay in the wilting process compared with that at room temperature.In addition,the chlorophyll content could not be used to correctly describe lettuce freshness because it showed a trend of increasing and then decreasing in different time periods.It can be seen that the hyperspectral imaging technology can realize the effective determination of lettuce freshness and water loss rate.

In view of the problem that the scene information is not be expressed accurately in the visible intensity image under some environment,a polarization image fusion algorithm based on residual dense blocks and attention mechanism is proposed in this paper.The proposed algorithm network consists of an encoder,a fusion module and a decoder.In the encoder,a residual dense block is constructed to preserve more feature information and enhance network stability.In the fusion module,channel attention mechanisms are embedded in the intensity feature map extraction network,while spatial attention mechanisms are embedded in the polarization degree feature map extraction network.The Sobel operator is employed to extract gradient information from shallow feature maps,which can enhance the detailed feature extraction ability of network and improve the utilization rate of feature maps.In the decoder,the feature maps in the encoder are skip-connected to corresponding convolution layers in the decoder to retain more feature information.Experimental results demonstrate that the fused images obtained by the proposed algorithm not only achieve the best values in multiple objective evaluation metrics,but also have better visual effects and more conform to the human visual perception.

In this paper,a novel polarization maintaining fiber （PMF） alignment method based on fiber loop mirror （FLM） has been proposed and developed.The two PMFs to be spliced are fused to the two arms of a 3 dB coupler and the other end faces of them are connected to build a FLM.The transmittance function of the FLM has been derived and the method to calculated the angle between the polarization axes of the two PMFs has been deduced.Finally,by analyzing the output spectrum of the FLM,the angle between the two PMFs can be calculated out and the two PMFs can be polarizationally aligned at any angle.The method has been analyzed in detail and verified by numerical simulation and experimental research.Aligning fibers based on the angle directly,this method is suitable to all kinds of fibers and has great application value..

Aiming at the characteristics of non-stationary,nonlinear,and susceptible to noise interference in distributed fiber optic sensing signals,as well as the problem of low recognition rate of submarine cable state by a single sensor,a multi-sensor fusion based method for identifying the exposed state of submarine cables is proposed.Firstly,the optical fiber sensing signal is processed using optimized variational mode decomposition (VMD),and the intrinsic mode function (IMF) is selected using the correlation coefficient method.Secondly,the IMF components selected by multi-sensor are sequentially arranged and encoded into grayscale images.Finally,a deep convolutional neural network (DCNN) structure is designed,inputting the training set into the network for training,and validating the effectiveness of the network with the test set to achieve recognition of the exposed state of submarine cables.By using on-site collected temperature and vibration data of submarine cables,the testing accuracy reaches 99.90%,and the results show that this method can accurately identify the exposed state of submarine cables;The testing accuracy of adding Gaussian noise to the original signal reaches 99.75%,proving that this method has good generalization ability and anti-noise performance.

To address the problem of accurately regulating the surface plasmon (SP) characteristics of metal-dielectric nanocomposites,a composite structure with a dielectric matching layer spacing was designed.The numerical relationships between the particle size and spacing of metal nanoparticles,the dielectric constant and thickness of the dielectric matching layer,and the near-field local enhancement and far-field scattering characteristics were studied by finite-difference time-domain method (FDTD).The results show that,in the incident light range of 400—900 nm,the resonance peak of the composite structure redshifts with the increase of gold nanoparticles and the decrease of particle spacing.The energy of the composite structure is highly concentrated in the dielectric matching layer,and when the dielectric constant is between 3 and 5,the coupling of localized surface plasmon (LSP) and propagation SP can be induced,which significantly enhances the electric field strength of SP.Meanwhile,the increase of the thickness of the dielectric matching layer can realize the redshift of the resonance peak.The results in this article verify the reliability of the method to comprehensively regulate the SP through the structure of metal nanoparticles and dielectric matching layer.

Laser polishing is a new surface treatment technology which can effectively improve the surface roughness of metal materials.In order to investigate the effect of laser polishing on the surface morphology of 316L stainless steel,in this paper,a two-dimensional axisymmetric transient laser polishing numerical model is developed to construct the model surfaces with high consistency with actual material surface contours,using a moving Gaussian pulsed laser as a heat source.The temperature distribution,velocity and the evolution of molten pool on the surface of the polishing model are simulated.The results show that when the surface curvature of the molten pool is large,the capillary force dominates the flow of the molten material from the crest to the trough; when the surface of the molten pool is smooth,the thermocapillary force dominates the flow of the molten material to the edges of both sides of the molten pool.The depth and duration of the molten pool have a great influence on the polishing effect,the surface roughness results are worse when the molten pool depth is small and the duration is short and when the molten pool depth is large and the duration is long.A set of laser parameters with better polishing results are analyzed and experimented,and it is found that the surface roughness of the material is reduced by 69.5% after polishing,which verifies that the numerical simulation model has a high accuracy.

With the increasing demand of access network bandwidth,100 Gb/s/λ coherent passive optical network (PON) has become the key of future PON.In this paper,a 100 Gb/s/λ homodyne coherent dual polarization-16 quadrature amplitude modulation (DP-16QAM) PON system based on blind phase search (BPS) algorithm is proposed.The system uses pulse-shaped polarization division multiplexing (PDM)16QAM signals on the optical line terminal (OLT) side to reduce the symbol rate.On the side of optical network unit (ONU),homodyne coherent detection technology is used to receive signals,and the BPS algorithm is used to process the system phase noise during the digital signal processing (DSP) process.The simulation results show that the system can achieve a power budget of 32.2 dB after 20 km standard single mode fiber (SSMF) transmission.At the same time,by using the BPS algorithm with a high linewidth tolerance,when the bit error rate (BER) is 10-4,after 20 km of fiber transmission,the sensitivity loss of the 10 MHz laser line width is only 0.6 dB more than that of 0.1 MHz,which allows the large linewidth laser to be used on the OLT side of the coherent PON system to reduce system costs.

To reduce the dispersion of optical signals in optical fibers and improve the output performance of the system,a cascaded fiber Bragg grating (FBG) structure is proposed in this paper.Firstly,three compensation schemes (pre-compensation,post-compensation,symmetric compensation) for dispersion compensation fiber (DCF) are analyzed,and it is concluded that the post-compensation scheme offers the optimal system performance.Next,the impact of post-compensation under different dispersion compensation methods on the system is analyzed.Subsequently,focusing on the post-compensation scheme,an FBG is cascaded after its transmitter to achieve a narrow optical signal linewidth and minimize the reflectivity side lobes,thereby reducing propagation delay and improving system performance.In both cases,with and without cascaded FBG,the influence of input power and transmission distance on the system is analyzed.The results show that with the DCF post-compensation scheme and the cascaded FBG,the transmission distance reaches 240 km,and the optimal input power is 21 dBm,resulting in an increase in the Q value from 37.152 75 to 60.552 1,representing a 63% improvement in system performance.This confirms that cascading FBG in the system can enhance the overall output performance of the system.This structure provides a solid foundation for future high-capacity,long-distance optical fiber transmission and holds promising development prospects.

In ultra-high speed coherent fiber optical communication systems,the improvement of single-carrier signal rate is usually considered as the foundation for rate improvement of wavelength division multiplexing (WDM) and even the entire communication system.During the process of single-carrier rate improvement,polarization mode dispersion (PMD) is considered to be one of the most important factors restricting the rate improvement of fiber optical communication systems.Conducting corresponding PMD simulation calculations is expected to provide PMD solutions for breakthrough rates on the one hand; On the other hand,it can lay a good foundation for subsequent Tb/s rate experiments and commercial use.This article quantitatively analyzes the impact of first-order,second-order and higher-order,full-order PMD on coherent fiber optical communication systems at the single-carrier rate of Tb/s level from the perspective of error vector magnitude (EVM) and symbol error rate (SER).Secondly,the constant modulus algorithm (CMA) is analyzed for first-order,second-order and higher-order,full-order PMD compensations.Finally,the optimal tap numbers of the CMA algorithm are calculated for three cases of PMD.The results show that the optical signal-to-noise ratio (OSNR) tolerance limit of the system can be improved by 6.7 dB and 6 dB,respectively,when only the first-order,second-order and higher-order PMD is included.When the single-carrier rate reaches 1.2 Tb/s,the compensation effect of the CMA algorithm for second-order and higher-order PMD is better than first-order PMD,and the optimal tap numbers are different in three cases.

Retinal vascular image segmentation is an important and difficult task in medical image analysis,and it is difficult for conventional methods to detect small and dense vascular structures effectively.To solve this problem,a novel high-precision and high-accuracy retinal vascular segmentation method that combines swin transformer block (STB) and full-scale attention skip connection technology (FSASC) is proposed.By constructing a U-shaped encoder-decoder network,the proposed method realizes self-attention from local to global,so that the proposed model can pay more attention to key vascular features.FSASC structure is used for fusing different features,which provides a simple and powerful mechanism for the proposed model to learn multi-scale semantic and spatial information.The proposed method is tested by using open datasets DRIVE and STARE.The experimental results show that the method can achieve high-quality and high-precision segmentation for retinal vascular structures.Compared with Unet and other methods,the proposed method has better performance in both the detailed feature segmentation and the segmentation accuracy.