Guided by improving the sensitivity of fiber Bragg grating (FBG) acceleration sensors,the sensitivity enhancement mechanism model is theoretically studied,and evaluation criteria is established for optimization and comprehensive performance.First of all,according to the working mechanism of inertial devices,a new general physical model of FBG acceleration sensor is established,the sensitive structure of the sensor and the design method of FBG are discussed and analyzed,and the response mechanism of its sensitivity is studied.Secondly,we analyze the influence of the inertial mass,the length of the encapsulated fiber,the stiffness of the sensitive structure and the Young′s modulus of the fiber on the sensitivity,and explore the threshold edge conditions of the stiffness of the sensitive structure to achieve high sensitivity.Finally,the concept of the "figure of merit" of the FBG acceleration sensor is proposed,combined with the threshold conditions of the sensor′s physical parameters,we study the relationship between the figure of merit and the inertial mass and fiber length.This not only provides a reliable basis for the evaluation of the comprehensive performance of the FBG acceleration sensor,but also has important theoretical guiding significance and practical application value for the design and optimization of the sensor.

Photonic crystals have become an important optical topology research platform in recent years due to their excellent properties such as designability,tunability,and supernormal control of light.The singular characteristic of Dirac cone linear dispersion has caused many interesting physical phenomena,such as Dirac oscillation,topological edge states,zero refractive index and so on,which are the physical roots of condensed matter topological phenomena.This paper focuses on the application of Dirac photonic crystals in surface emission lasers in recent years, and points out that the introduction of Dirac photonic crystals into semiconductor lasers can realize large area ultra-low threshold, high brightness,single longitudinal mode and single transverse mode topological cavity surface emission lasers (TCSELs),and also summarize and look forward to the TCSELs developed based on the Dirac photonic crystal principle.

In order to solve the problem that the backscattered signal is seriously disturbed by noise in an optical time domain reflectometer (OTDR),this work proposes an OTDR signal denoising algorithm based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) improved wavelet threshold.By using the CEEMDAN decomposition algorithm to resist modal aliasing and reduce reconstruction errors,the signal is decomposed into several intrinsic mode function (IMF) components.On the basis of the analysis method of the correlation coefficient,the critical point between the noise-dominated IMF components and the signal-dominated IMF components is found,and the noise-dominated IMF components are removed.Then the signal-dominated IMF components are denoised by the improved wavelet threshold denoising method,and the signal is finally reconstructed.The results show that the proposed method can suppress the noise better and achieve better results and highlight the event features compared with the traditional hard threshold method, CEEMDAN-hard threshold method and the improved wavelet threshold method and make event detection easier.

Under the standard test conditions,the spectral form,the irradiance intensity and the spectral mismatch factor of artificial solar light sources are important factors that affect the accuracy of solar cell device test results.This paper takes the standardsolar spectrum AM1.5 as areference,then calculates and analyzes the changes of thespectral mismatch factor between the four commonly used artificial solar light sources (Arc lamp,Q-Flash lamp,Q-Flash W lamp and ELH lamp) and thestandard solar spectrum AM1.5 from the perspective of spectral mismatch,and also calculates and analyzes the changes in the output parameters of the crystalline silicon cell under the irradiation of these four artificial solar light sources.Calculation results show that the spectral mismatch factor of Arc lamp artificial solar light source is 0.979,and the output parameters of the crystalline silicon cell are closest to the output parameters under the standard solar spectrum AM1.5 spectral irradation;due to the influence of spectral mismatch,the output parameters of the crystalline silicon cell under the irradiation of four different artificial solar light sources will change significantly compared with the output parameters under the irradiation of the standardsolar spectrum AM1.5 spectrum.

A magnetic field detection micro-cavity model based on parity-time (PT) symmetry structure is proposed.Using the magnetic-optical effect of hydro-based Fe3O4 magnetic fluid medium layer and the optical amplification effect of the PT-symmetry structure,when the magnetic field strength changes,the defect mode position in the structure band gap is changed.Therefore,the magnetic field intensity can be detected by detecting the defect mode position.In this paper,the transmission spectrum of the structure is theoretically analyzed based on the transfer matrix method (TMM),and the parameters such as the structure period,the thickness of the dielectric layer and the macroscopic Lorentz oscillation intensity in the PT-symmetry element are numerically optimized.The optimized simulation results show that the sensor can achieve effective magnetic field intensity detection,the sensitivity of the sensor can reach 97.14 nm/RIU in the magnetic field intensity range of 0—500 Oe,the figure of merit and detection limit can reach 104 and 10－7 RIU,respectively.Meanwhile,the resolution value can be as low as 0.05 Oe and as high as 2 Oe.The proposed sensor structure can be used in the design of high comprehensive performance magnetic field intensity testing equipment.

Convolutional neural networks show strong feature learning ability in advanced computer vision and have achieved remarkable effect in image semantic segmentation tasks.However,how to use the multi-scale feature information effectively is always a difficulty.This paper proposes an effective image semantic segmentation method which integrates multi-scale features.The proposed method consists of four basic modules,which are feature fusion module (FFM),spatial information module (SIM),global pooling module (GPM) and boundary refinement module (BRM).FFM adopts attention mechanism and residual structure to improve the efficiency of multi-scale feature fusion.SIM includes convolution and average pooling opearaitons,and its purpose is to assist in locating the edge information of the object by providing additional spatial details.GPM extracts the global information of the image,which can significantly improve the performance of the model.BRM takes the residual structure as the core to refine the boundary of the feature map.Four basic modules are added into the full convolutional neural network to effectively utilize the multi-scale feature information.Experimental results on PASCAL VOC 2012 dataset show that mean intersection over union of the proposed method is 8.7% higher than that of full convolutional neural network.The results of comparison with other methods in the same framework also verify the effectiveness of the proposed method.2021-07-05

The morphological structure of retinal vessels is an important index to reflect human health.In order to solve the existing problems in retinal vessel segmentation,such as blurred main vessels,broken microvessels and false segmentation of optic disc,the multi-scale feature fusion double U-shaped retinal segmentation algorithm is proposed.Firstly,the low level U-Net efficient cyclic residual module is used for coarse-grained segmentation of fundus images to obtain the initial contour of retinal vessels.Secondly,the coarse segmentation image is multiplied by the pixels of the original feature image into the high level U-Net,and the scaling wide residual model is used to decode the fine-grained image to enrich the details of retinal vessels.At the sametime,the three pathway attention mechanism is used to connect the encoding layer and decoding layer of the double network in a compound way to realize the cross network propagation of feature mapping and reduce the semantic difference of context.Finally,the double U-shaped network can extract vascular pixels at a deeper level and accurately segment retinal details.Experiments were conducted on DRIVE and STARE datasets,the accuracy was 96.45% and 97.02%,the sensitivity was 83.35% and 81.40%,and the specificity was 98.38% and 98.83%,respectively.The overall performance is better than existing algorithms.

In order to further improving the performance of infrared and visible image fusion method,an infrared and visible image fusion method based on multiscale local extrema decomposition (MLED) and deep learning network ResNet152 is proposed in this paper.Firstly,the source images are decomposed into approximate images and many detail images using MLED,which can separate out the overlapped important feature information.Secondly,the residual network ResNet152 is used to extract the multi-dimensional deep features of the source images,and the ｌ1-norm is used as the activity level measure to generate the salient feature maps,the weighted average fusion algorithm is carried out for the approximate images,which can keep the energy and residual details not lost.For the detail images,the initial decision map is obtained by the rule “coefficient absolute max”.The source images are used as the guided images,and the initial decision maps are used as the input images for guided filtering.So the optimized decision maps are obtained,the weighted local energy is calculated to get the energy saliency maps.The weighted average algorithm is carried out for the detail images,which can make the fusion image having rich texture details and good visual edge perception.Finally,the fusion image is obtained by reconstructing the fused approximate image and detail images.The experimental results show that,compared with the existing typical fusion methods,the proposed method can achieve state-of-the-art results in terms of both objective evaluation and visual quality.

This paper starts from the development trend and research background of ultraviolet (UV) optical communication technology,for the application requirements of target localization in specific environments,a method based on single scattering of polarized ultraviolet light to achieve non-line-of-sight target localization is proposed by combining the vibration direction characteristics of polarized light.Firstly,based on the UV atmospheric scattering transmission characteristics,a polarized UV non-line-of-sight single scattering model is established,and the polarization scattering propagation characteristics of ultraviolet light is derived by the method of matrix optics.The relationship between the received light intensity and azimuth angle and distance between transmitter and receiver is established.Then,the Matlab software is used to simulate and analyze the effect of transmitter and receiver elevation angle and field of view angle on the received light intensity and azimuth angle measurement range,and the suitable transmitter and receiver elevation angle and field of view angle are selected.The azimuth angle of the target is obtained according to the change of the light intensity of the received signal with the azimuth,and the distance between transmitter and receiver is calculated.In theory,the feasibility of the system is verified,and the simulations and analyses of each error influence are carried out.At a distance of 500 m,the distance measurement error is 25.3 m when the transmitter and receiver elevation angle deviates 3° (preset value 25°).The influence of noise on the light intensity leads to the azimuth angle determination error to have a small effect on the ranging error,but the azimuth angle determination error of 3° can cause a lateral deviation error of 26.2 m from the true position at a distance of 500 m,thus improvement measures are proposed.The results of this paper provide a theoretical basis for expanding the functions and practical applications of UV optical communication systems and have certainguiding significance.

In a hybrid optoelectronic data center,an electronic switch can switch packets in small traffic flow flexibly and efficiently,but cannot effectively transmit those in large traffic with long duration.In contrast,optical routers are just the opposite.Therefore,in the hybrid optoelectronic data center networks based on arrayed waveguide grating router (AWGR),a hierarchical and classified traffic scheduling strategy using active queue detection is proposed to achieve effective scheduling of both large and small traffic by a combination of electronic switching networks and optical wavelength paths.According to the traffic model of data centers,simulation experiments are designed,in which the average delay and throughput of each architecture under different traffic models are analyzed.Numerical results show that:comparing with the Fat-tree,C-Through structures and the conventional tree structures with a bandwidth convergence ratio of 1∶1,the throughput of the AWGR structures increases at least 19.43% and the delay reduces at least 45.73% under random traffic model.Under point-to-point traffic model,as K increases from 1 to 4,the AWGR structures compared with the other three structures have the lowest throughput drop and delay increase of 2.85% and 46.98% respectively,Our scheme can effectively improve the performance of throughput and delay in a datacenter.

In order to realize the strict dual-mode transmission and large mode field area of the multi-core optical fiber,the multi-core fiber is introduced into the air hole and designed into an octagonal structure.First,The electromagnetic field model of the fiber by using COMSOL software is established,and then the finite element method is used to systematically analyze the influence of the three structural parameters of the relative aperture size,the core-clad refractive index difference and the core spacing on the fiber mode characteristics and the effective mode field area.The bending loss of the optical fiber is discussed under different bending radii in the end.According to the analysis results and combined with the normalized frequency constant to find suitable structural parameters,the fundamental mode field area of the fiber can reach 1 〖ＫＧ－１／６〗730 μm2 in the flat state.When the bending radius is greater than 0.45 m,the bending loss is less than 10－3 dB/ m,the area of the fundamental mode field can still reach 1 685 μm2 at this time.This optical fiber maintains few-mode transmission and achieves a large mode field area and low bending loss,so it has broad application prospects in large-capacity and high-power optical fiber transmission systems.

Based on the generalized Huygens-Fresnel principle and the theory of coherence and polarization of light,it was derived that the analytical expression of the cross spectral density matrix element of random electromagnetic Gaussian vortex beam in gradient-index fiber,the spectral degree of coherence of the random electromagnetic Gaussian vortex beam propagating through gradient-index fiber is studied.The results show that when random electromagnetic Gaussian vortex beam propagating in gradient-index fiber,the period of the spectral degree of coherence is related to the gradient-index coefficient β.the longer beam wavelength is,the more obvious the focusing effect of the gradient-index fiber on the distribution of the spectral degree of coherence is.The number of dark rings in the distribution of the spectral degree of coherence is the same as the topological charge m.The value of spatial correlation length σxx change the evolutive speed of the spectral degree of coherence along the radial direction,while spatial correlation length σxy dose not change the distribution of the spectral degree of coherence.The research results of this paper have guiding significance for the propagation of random electromagnetic vortex beams in optical fibers,and have reference value in high-speed and large-capacity optical fiber secure communication.

In view of the fact that the current robust watermarking is difficult to resist both conventional attacks and geometric attacks,a new contourlet watermarking algorithm is proposed.In this watermark embedding stage,the host image is transformed by discrete wavelet transform and non-subsampled contourlet transform,and the watermark is embedded into the high frequency subband with the maximum entropy,which solves the problems that embedding the watermark in the low frequency can not resist geometric attacks and high frequency can not resist conventional attacks such as filtering.The embedded area is encrypted before the watermark is embedded,which improves the security of the algorithm.In the watermark extraction stage,the watermark is geometrically corrected and extracted by using Zernike moments,which further improves the anti-geometric attack ability of the algorithm.Finally,the algorithm is evaluated by peak signal-to-noise ratio and normalized coefficient to verify the imperceptibility and robustness of the algorithm.Simulation results show that the algorithm performs well on different host images and watermark images.On the basis of good invisibility,the algorithm not only has good resistance to conventional attacks such as filtering and noise,but also has strong robustness to geometric attacks such as translation and rotation,and the normalization coefficient (NC) value is more than 0.99.