In order to further improve the frequency mutiplication coefficient of photogenerated millimeter-wave in the radio over fiber (ROF) system,a scheme for generating frequency 20-tupling millimeter-wave signals based on the interaction of four Mach-Zehnder modulators (MZM) is proposed.The generation mechanism of frequency doubling scheme under ideal conditions is deduced.In the simulation experiment,the influences of DC bias voltage drift,modulation index and extinction ratio on the system performance under non-ideal factors are analyzed respectively.The results show that the saturation values of the optical sideband suppression ratio (OSSR) and the radio frequency spurious sideband suppression ratio (RFSSR) can reach 33.20 dB and 27.21 dB when the range of each parameter is set properly.For the ROF system based on this frequency doubling scheme,two different transmission modes of 2.5 Gbit/s data signal with single tone and dual tone modulation formats are compared and analyzed.The simulation results show that when the optical fiber distance is 40 km,the single sideband can still achieve error-free transmission,which reduces the influence of inter-code walk-off effect in the transmission process,increases the transmission distance of the system,and is more suitable for long-distance transmission.It provides a theoretical basis for the development of microwave photonics.

Michelson interferometer sensor based on cylindrical vector modes (CVMs) in a vortex fiber (VF) has been proposed and demonstrated.The sensing characteristics of axial strain,temperature,and refractive index have been studied using theoretical and experimental methods.By applying a microbend long period grating (MLPG) to the VF,the first-order CVMs of TE01,HE21 and TM01 modes are excited.The end of the VF is equipped with a mirror,which interferes with the fundamental mode HE11 in the fiber core and the CVMs in the ring core after reflection,constructing an in-fiber Michelson interferometer sensor.The sensor exhibits the strain sensitivity of -1.01 nm/mε, -1.61 nm/mε, -2.31 nm/mε and the temperature sensitivity of 9.3 pm /℃, 3.4 pm /℃, -2.4 pm /℃ for the CVMs of TM01,HE21 and TE01 respectively.But it is insensitive to refractive index.The sensor has a simple structure.The TE01 mode has the highest strain sensitivity and the lowest temperature sensitivity.It can be used as an ideal uncompensated strain sensor with insensitive refractive index and low temperature cross sensitivity.It has potential application of the next-generation sensors in smart engineering structures.

Aiming at the research of nondestructive and non-contact measurement based on optical coherence,the method of designing and constructing common optical path laser shear interference system is proposed using Wollaston prism (WP) as the shear component.In this paper,the effects of its components on the imaging quality of interference fringes are analyzed in detail,and a series of conditions to obtain the optimal interference fringe are acquired.Subsequently,the convective vortex cell structure and temperature distribution of the fluid in the transparent square cavity under the action of temperature gradient are obtained by using the constructed system,which preliminarily verifies the feasibility of the system in this research field.On the basis of the above research,combined with image processing and iterative algorithm technology,the system has a great application prospect in three-dimension field reconstruction.

To solve the problem of low sensitivity and temperature cross-sensitivity of optical fiber transverse load sensor,an elliptical structure optical fiber transverse load sensor with temperature self-compensation is proposed.The sensor is formed by bending a standard single mode fiber (SMF) into an elliptical structure and encapsulated in polydimethylsiloxane (PDMS).The temperature self-compensation is realized by paralleling another nearly twin elliptical structure of the measuring element.The core mode and cladding mode interfere in the elliptical bending SMF,and the interference spectrum shifts with the change of applied transverse load.The experimental results show that the characteristic wavelengths of the sensor have a linear relationship to the transverse load in the range of 0.25—0.5 N.In the temperature range of 33.5—44 ℃,the sensitivity after temperature self-compensation can reach 6.6 nm/N,the temperature cross sensitivity is only 0.001 5 N/℃,and the temperature self-compensation error is less than 0.089 nm.The proposed transverse load sensor has high sensitivity, low cost and simple structure.It has certain application reference value.

The terahertz (THz) slow-light effect can effectively improve the security and memory property of THz pulse data during transmission. However,the general THz slow-light devices are sensitive to the polarization state changing of the incident THz wave.In this paper,we design a metamaterial structure whose unit cell consists of one cross resonator and four U-shaped resonators.The results show that the THz slow-light effects based on metamaterials are not sensitive to both linear and circularly polarized light. A slow-light effect with the maximum group refractive index of 1 618 is obtained by optimizing parameters of the metamaterial structure.A molybdenum disulfide film between the metamaterial microstructure layer and the SiO2 substrate is embedded.When the carrier concentration of molybdenum disulfide increases from 1.7×1017 cm-3 to 5.1×1019 cm-3,the group refractive index decreases from 1 566 to 26,and a polarization insensitive all-light tunable THz slow-light effect is realized.This study can provides some new ideas for the design of polarization insensitive and all-light tunable THz slow-light devices.

Tampering with joint photographic experts group (JPEG) images often produces double JPEG (DJPEG) compression traces,and analyzing the traces can help reveal the image compression history and enable tampering region localization.Existing algorithms perform poorly when the image size is small and the quality factor (QF) is low, and there are restrictions on the combination of the two QFs.In this paper,an end-to-end mixed QF DJPEG compressed image forensics network named DJPEGNet is proposed.First,the preprocessing layer is used to extract the quantization table (Qtable) features representing the compression history information from the image header file,and the image is converted from the spatial domain to the discrete cosine transform (DCT) domain to construct statistical histogram features.Then,the two features are input into the main structure formed by stacking the depthwise separable convolution and residual structure,and the binary classification result is output.Finally,a sliding window algorithm is used to automatically locate the tampered region and draw a probability distribution map.The experimental results show that, on small-size datasets generated by different Qtable sets,DJPEGNet outperforms the existing state-of-the-art algorithms in all indicators,with ACC increased by 1.78%,TPR increased by 2.00%,TNR increased by 1.60%.

GF-3 satellite is the first C-band multi-polarimetric synthetic aperture radar satellite with a space resolution up to 1 m in China,in which scan synthetic aperture radar (ScanSAR) is one of the important mode of GF-3.The working mechanism of this mode results in the phenomena of serious nonuniformity,generally showing bright and dark stripes,also known as scalloping.In view of scalloping in ScanSAR mode of GF-3,this paper proposes a model combining self-attention mechanism and cycle-consistent adversarial networks (CycleGAN),so as to perform descalloping.The proposed descalloping method is compared with traditional descalloping methods and deep learning related algorithms,and analyzed by indicators such as brightness average and average gradient.The experimental results demonstrate that the proposed method in this paper can better complete descalloping in the GF-3 ScanSAR image,effectively suppress the stripes phenomenon of the image,and improve the image quality,which is of great practical significance.

In stereo matching of binocular vision, due to the lack of texture information on the same-tone surface,not only has a large amount of computation but also has low matching degree,and the point cloud in the generated scene has the nature of unstructured,the near is dense and the far is sparse.Therefore,improving the matching accuracy and speed of binocular vision,and accurate segmentation of the target has been a difficult problem in point cloud acquisition and target detection.To solve the above problems,firstly,a 3D point cloud target acquisition method combining active laser is proposed to obtain the original point cloud data quickly and accurately.And an improved algorithm based on Euclidean clustering is proposed,which uses distance threshold and angle threshold as the threshold segmentation judgment conditions to perform segmented clustering,the 3D point cloud target detection box with clear boundary is obtained. The experimental results show that the designed 3D point cloud imaging system can effectively obtain the 3D point cloud information of the target in front, and has the advantages of lower cost, easier implementation and more information than lidar.The improved Euclidean clustering algorithm can effectively solve the problem that the object is prone to under-segmentation or over-segmentation,because the traditional algorithm is sensitive to the threshold,and the accuracy of target detection is improved,the detection effect is better in indoor scenes.

In the action recognition task,how to fully learn and utilize the correlation between the spatial features and temporal features of the video is particularly important for the final recognition results.Aiming at the problem that the traditional action recognition method ignores the correlation of spatio-temporal features and small features,which leads to the decrease of recognition accuracy,this paper proposes a human action recognition method based on convolutional GRU (ConvGRU) and attentional feature fusion (AFF).Firstly,the Xception network is used to obtain the spatial feature extraction network of video frames,and the spatio-temporal excitation (STE) module and channel excitation (CE) module are introduced to obtain the spatial features and strengthen the modeling ability of temporal actions.In addition,the traditional long short term memory (LSTM) network is replaced by the ConvGRU network,which uses convolution to further mine the spatial features of video frames while extracting temporal features.Finally,the output classifier is improved,and the feature fusion module based on improved multi-scale channel attention is introduced to strengthen the recognition ability of small features and improve the accuracy of the model.The experimental results show that the recognition accuracy of 95.66 % and 69.82 % are achieved on the UCF101 dataset and the HMDB51 dataset,respectively.The algorithm obtains more complete spatio-temporal features and is superior to the current mainstream models.

Isotopes play an important role in agriculture,medicine,military weapons,energy and environment,et al.so the detection of isotopes is particularly important.A new method of isotope identification based on reaction microscope spectroscopy is presented in this paper.The nitrogen and oxygen are the basic elements that make up the main components of air and they play an important role in all kinds of life activities.Because nitrogen and oxygen have stable isotopes and significant isotopic effects in nature,the determination of their isotopes is very important. In this paper,nitrogen and oxygen are selected as target gases to carry out electron collision experiments with the help of the third-generation reaction spectrometer,namely the reaction microscopic imaging spectrometer.Isotopes in the target gases of nitrogen and oxygen are successfully determined,and the same isotope abundance ratio as in nature is obtained in the results,which verifies the reliability of isotope determination by the reaction microscopic imaging spectrometer.The ability to distinguish isotopes with only one difference in mass fraction also demonstrates the extremely high measurement accuracy of current reaction spectrometer devices.

The fractal study of the lightning time domain waveform ignores its frequency characteristic,so that all the characteristics can not be fully characterized.In order to solve this problem,this paper introduces the multifractal theory into modern spectral estimation,and proposes a multifractal characteristic analysis and discharge type identification method of the lightning electric field signal based on auto-regressive (AR) spectrum.Firstly,the power spectrum of the lightning electric field signal is obtained based on the AR model spectrum estimation method.Then the multifractal detrended fluctuation analysis (MF-DFA) method is used to verify that the lightning AR spectrum sequence has multifractal characteristics,and the Hurst exponent and multifractal spectrum of AR spectrum sequence are further discussed.Finally,these parameters are input into support vector machine as the effective eigenvalues to identify different discharge types of intracloud lightning (IC) and cloud-to-ground lightning (CG).The experimental results show that the effective recognition rate of the proposed method reaches more than 94%.The research results have certain reference value for the research of lightning characteristics and automatic recognition technology.

The spectral shifts and spectral switches of partially coherent radially polarized Laguerre-Gaussian (PCRPLG) pulsed beams propagating through atmospheric turbulence are investigated.It is found that the on-axis spectrum of the beams will shift rapidly in atmospheric turbulence due to the appearance of two spectral peaks caused by the turbulence.The stronger the turbulence is,the more stable the relative spectral shifts are after the rapid shift, and the off-axis spectrum has a smaller variation range of relative spectral shifts in stronger turbulence.It is also found that the influence of atmospheric turbulence on the spectral shifts of the on-axis and off-axis spectra can be significantly reduced by appropriately increasing the pulse width.Under different pulse widths,spectral switches appear in both on-axis and off-axis spectra,and the off-axis spectrum will produce a unique phenomenon of two rapid shifts.For a fixed transmission distance,the higher the topological charges are,the smaller the spectral shift of the on-axis spectrum is,and the smaller the variation range of the relative spectral shifts in the paraxial region is.In addition,spectral switches will appear in different topological charges.The larger the coherent length is,the more stable the spectral shifts are, and the farther the spectral switch occurs in the off-axis spectrum.The results of this paper provide a theoretical basis for the application of Doppler lidar.The PCRPLG pulsed beam itself has a strong anti-turbulence ability, and the spectral shifts of the beams can be reduced by adjusting the beam parameters.

Aim at traditional contactless true temperature inversion algorithm has the problems of slow inversion speed and low accuracy,a new constrained true temperature inversion algorithm (CTTIA) based on luminance temperature model is proposed.In the whole modeling process,it is found that emissivity and luminance temperature have internal relations:the universal law from luminance temperature to emissivity or from emissivity to luminance temperature.The simulation results show that when the number of emissivity samples is huge,the CTTIA can not only provide theoretical guidance for the experiment,but also greatly improve the selection efficiency of emissivity samples.Nine wavelength channels are built to measure and calculate the measured objectives at 1 800 ℃ and 2 000 ℃.The results show that the accuracy of the CTTIA is almost the same as that of the second measurement method (SMM),and the inversion time is reduced to 82%.It illustrates that the research of this method is very critical and important,has great research value.

An improved 3D UNet network model that combines a dual path attention (DPA) module and a multi-scale feature aggregation (MFA) module is proposed to address the problems such as blurred edges,low contrast,and uneven gray value distribution in magnetic resonance (MR) images of the prostate bringing about the poor segmentation accuracy.Firstly,the dataset is resampled and cropped to fit the model input.Then,DPA and residual connection are added to each layer of the 3D UNet network encoder to enhance the feature coding capability.At the same time,an MFA module is added to the network decoder to make full use of spatial context information and enhance semantic information.Finally,the proposed model is validated on the public dataset PROMISE12,with the Dice coefficient of 89.90% and the Hausdorff distance of 9.37 mm. Compared with other models,the proposed model has better segmentation results,and the number of parameters and the amount of computation are less.