Hollow core terahertz (THz) fibers have attracted attention due to the low loss and easy modification with functional materials.Liquid water has unique properties in the THz region and has been widely investigated in THz emission,sensing and devices.In this paper,a THz hollow core Bragg fiber (HCBF) with a water defect layer is proposed.The finite element method is used to analyze the temperature characteristic of the water-filled THz fiber.The numerical analysis results show that the peak of the confinement loss can be controlled by adjusting the temperature of the liquid water,and the confinement loss increases with the increasing temperature of the liquid water in the frequency range of 0.34 THz to 0.44 THz.Moreover,the THz fiber has a high core power fraction up to 98.6%.The proposed THz water-filled fiber may have great potential as a sample cell for THz interaction with liquid and THz controllable devices.

Based on the key parameters,such as X-ray energy spectrum range,source position and imaging distance,a multi-cone crystal spectrometer with a wide spectral range the and high spectral resolution was designed to overcome problems of aberration detection in the traditional curved crystal spectrometer.The characteristics of the spectrometer are discussed in terms of structure and principle.A continuous and smooth spectrometer has been developed.By single X-ray focal diagnosis,the aberration free detection of X-ray with different energy can be realized.The intensity distribution of the Kα and 〖ＷＴＢＸ〗Kβ spectra at different focusing imaging positions was obtained by multiple diffraction experiments,and the focusing ability and spectral resolution of the spectrometer were analyzed.According to the relationship between X-ray focusing degree and imaging position,the focusing performance of continuous cone spectrometer at different focusing imaging position is analyzed,and the best focusing position is determined.The experimental results show that the focusing ability of the multi-cone crystal X-ray spectrometer has been significantly improved,and the energy spectrum detection range can cover 4.51—5.14 keV,and the spectral resolution can reach more than 600 on the larger size of the light source.

A dual-parameter fiber-optic sensor based on Fabry-Perot interferometer (FPI) structure is described and demonstrated to measure gas pressure and temperature simultaneously.The proposed sensing structure consists of an extrinsic FPI in the form of hemispherical ultraviolet curing glue capped on a single mode fiber end face.The spectral interference peak produced by FPI is sensitive to ambient temperature and gas pressure variations.We select two resonant peak dips in the interference spectrum to track their wavelength changes with gas pressure or temperature,constructing transport matrices capable of measuring both gas pressure and temperature and eliminating mutual cross-sensitivities.The experimental results showed that the gas pressure and temperature sensitivity at dip 1 were of 2.925 nm/MPa and －1.165 nm/℃,and the gas pressure and temperature sensitivity at dip 2 were of 4.715 nm/MPa and －1.209 nm/℃,respectively.Building the measurement matrix using the gas pressure and temperature sensitivity of dip 1 and dip 2,the sensor can realize the multi-parameters measurement.The stability experiment of the sensor showed that the measurement error of the gas pressure and temperature were of 0.068 MPa and 0.19 ℃,respectively.The sensor has high sensitivity,low cost,simple and compact structure.It has certain application value in industrial production.

With the popularity of optical communication technology,there is a growing recognition that the filter plays a significant role in wavelength division multiplexing technology.However,traditional filters have some disadvantages which are supposed to be improved and optimized,such as large size,poor filtering performance and complicated preparation.In order to solve these problems effectively,this paper designed two types of one-dimensional photonic crystal narrow-band filters with (AB)NBM(AB)N and (AB)N(BA)N structures.At the same time,the transfer matrix method was be used to calculate two photonic crystal filters.The photonic band gaps were analyzed and the effect of different parameters on the filtering performance was explored on the basis of the optimized design structure.The results show that: the one-dimensional laminated defect type photonic crystal narrow-band filter and the one-dimensional mirror image type photonic crystal narrow-band filter are able to obtain a narrow transmission band within a certain wavelength range,the transmittance at the narrow band of laminated defect type photonic crystal is close to 99% and the mirror image type photonic crystal′s transmittance is 96%,as well as present excellent filtering performance,which can meet the narrowband filtering technical requirements.It is expected to be widely used in optical wavelength division multiplexing technology.

With the continuous development of three-dimensional (3D) data acquisition technology in recent years,the acquisition of three-dimensional large-scale scene point cloud data is becoming more and more convenient.At present,the deep learning network framework is becoming more and more mature in the field of two-dimensional (2D) image processing,while the large-scale scene point cloud is a kind of 3D irregular data.When using 3D convolutional neural networks in deep learning to directly process large-scale scene 3D data,there will be problems such as low classification accuracy and high computational complexity.Therefore,to effectively solve the problems of long computation time and low classification accuracy in point cloud classification based on deep learning,this paper proposes a binary neural network-based classification method for large-scale scene point cloud.designing the eigenvalue calculation method for irregular 3D point cloud data,processing the input point cloud feature images based on IR-Net binary neural network,further adopting Dynamic ReLU activation function to improve the computational efficiency of the neural network,and finally deriving the point cloud classification results.The experimental results show that the proposed method achieves 97.6% classification accuracy on the Oakland dataset and 92.3% and 97.2% in the GML dataset,Experimental results show that Dy-ResNet can effectively improve the accuracy of point cloud classification,reduce the complexity of calculation and improve training efficiency.

Aiming at the problems of color distortion and low contrast of underwater degraded images in different scenes,an underwater image enhancement method combining multi-channel equalization and multi-scale fusion is proposed.Firstly,the green channel is used to compensate the red channel and then the red channel is strengthened again through the regional red channel mean.Secondly,the self-adaptive platform histogram equalization is used to expand the gray-level range of R,G and B channels and redistribute the gray values.Finally,the V channel processed by Gamma,GUM and CLAHE algorithms is fused at multiple scales.The experimental results show that the proposed algorithm is simple and effective,which can significantly improve the haze and blue-green appearance of the image,enhance the image definition,and show strong generalization adaptability in a variety of scenes.

To address the problem of MRI brain tumor region misidentification and spatial information loss of segmentation network,an MRI brain tumor image segmentation method based on dual-branch feature fusion is proposed.First,the contextual information of the network is extracted by structurally the re-parameterization visual geometry group and attention model (RVAM) in the primary branch,and then the rich spatial information is obtained in the secondary branch using deformable convolution and pyramid pooling model (DCPM),after which the feature fusion module is used to fuse the feature information of the two branches.Finally,the attention model is introduced to strengthen the weight of segmented targets in the up-sampling process at decoding.The proposed method has been experimentally validated on the Kaggle~~3m and BraTS2019 datasets,and the experimental results show that our method has good brain tumor segmentation performance,where the Dice similarity coefficient and Jaccard coefficient reach 91.45% and 85.19% on Kaggle~~3m,respectively.

Image inpainting is one of an activate research topic in the domain of computer vision and computer graphics.Aiming at the problem that the traditional crack image restoration method using one-time completion restoration method does not have the ability to understand semantics,and the repair effect is not good when the semantic scene is more complex and the image defect is large,a crack image restoration based on progressive feature reasoning is proposed.This method gradually restores the image from the hole edge and strengthens the constraint on the hole center.At first,partial convolution is used to update the mask,and the update ratio is determined by calculating the mask proportion.Then,use the VGG-16 network for feature extraction,semantic attention mechanism is used to generate high-quality image features,and use the jump connection method to enhance the gradient correlation of remote distances,so as to provide multi-scale and multilevel feature information for subsequent image restoration.Finally,the recursive feature map is fused and decoded to generate a repair image.The experimental results show that the proposed method,compared with traditional image inpainting methods,can improved the peak signal-to-noise ratio of the crack image repaired for 0.5 dB—1.2 dB and produce semantic clear inpainting results.

Aiming at the problems of weak anti-interference ability and poor speed of dynamic foreign object detection in complex railway environment,a rail foreign object intrusion detection method based on an adaptive Gaussian mixture model (AGMM) is put forward in this paper.By analyzing the characteristics of randomness when compound jitter occurs in railway scenes,Firstly,jitter detection on the input railway video is performed,and then affine transformation and median filtering are introduced to process the jittery images in the video sequence.Secondly,the method of iterative filling frame by frame is used to fill the black edges of the image after debounce to obtain a railway video frame without jitter and without black edges.Finally,on the basis of the existing Gaussian mixture model,an adaptive selection of the number of Gaussian distributions and learning rate is designed,and the improved Gaussian mixture model is used to realize the background modeling of complex railway videos,and thereby improve the detection speed of foreground objects.The experimental results show that in the case of jitter in the railway video,the accuracy rate of the track foreign body intrusion target detection is 2.6 times,and the detection speed is 2.8 times that of the original algorithm,which can improve the anti-interference and rapidity of target detection.

Traditional spectrum sensing methods are susceptible to noise fluctuations and have poor detection accuracy under wireless communication conditions with low signal-to-noise ratio.The anti-noise performance and detection accuracy of traditional spectrum sensing methods can be effectively improved by combining the lifting wavelet denoising and dynamic threshold energy detection algorithm.First,perform odd and even sampling of the noisy signal,decompose the noise signal,remove the noise part,and reconstruct it into a new signal without noise,then use the energy detection method to count the energy accumulation of the signal,set the dynamic threshold,and finally use the dynamic threshold to determine whether the primary user signal exists.Lifting wavelet denoising can effectively remove the noise in the sampled signal and reduce the influence of noise on the detection accuracy of the energy detection method.The dynamic threshold can be adjusted according to noise fluctuations to adapt to the complex noise environment.The simulation results show that the lifting wavelet denoising combined with dynamic threshold energy detection algorithm has better detection accuracy than traditional spectrum sensing methods.This method improves its resistance to uncertain noise,enables it to adapt to complex communication environment,and also improves the reliability of spectrum sensing process.

Aiming at the problem of low accuracy of impurity detection in raw cotton,an improved algorithm based on residual and attention mechanism for detecting raw cotton impurity in Xinjiang cotton is proposed.The algorithm has high accuracy and is a two-stage algorithm.Firstly,the impurity images of raw cotton were collected and labeled,and then the data were enlarged to avoid the overfitting phenomenon in the training process.Then,visual attention mechanism is introduced into the original framework,and the accuracy of impurity detection of raw cotton is improved by advancing the algorithm structure.Secondly,by analyzing and comparing the accuracy of several different networks in detecting raw cotton impurities,ResNet50 was selected as the feature extraction network,which improved the complex feature extraction ability of the algorithm.Finally,ROI Align is used to reduce quantization errors and improve the accuracy of the detection of raw cotton impurities.Experimental results show that although the improved algorithm slightly increases the detection time,its overall detection accuracy is significantly better than the original algorithm,and the overall recognition accuracy can reach about 94.84%,which is 5.58% higher than the recognition rate of the faster region-based convolutional neural network (Faster R-CNN) before the improvement.Meanwhile,by comparing different network models,the results show that the improved Faster R-CNN has a better effect on the detection of raw cotton impurities.

In order to solve the tedious problems of traditional manual contact measurement with large workload,relatively difficult measurement conditions and affecting the normal living habits of cattle,a method for measuring cattle body size based on binocular depth estimation with target detection is proposed.Firstly,the images are collected by a binocular camera,and the object detection algorithm YOLOv5 (you only look once v5) is used to detect the feature parts of the body scale in the images,and combined with edge detection and other algorithms to obtain the cattle body scale measurement points.Secondly,the binocular stereo matching algorithm is used to transform the binocular two-dimensional images into a spatial three-dimensional depth information map,and the three-dimensional coordinates of the measurement points are read on the depth information map.Finally,the spatial Euclidean distance is used to calculate parameters of the cattle body scale in three-dimensional coordinate system.A test platform is built for the measurement,and the experimental results show that the measurement accuracy of the method is higher than the existing related methods,in which the average relative error of body length is 2.4%,the average relative error of body height is 5.9%,and the average relative error of the body oblique length is 3.3%,which provide a feasible demonstration for the measurement of cattle body scale.

In order to meet the requirements of the sixth generation mobile communication (6G) system for high speed and large capacity of optical communication network,further improving the bandwidth,response rate and amplification of optical fiber amplifier in optical transmission network has become the focus of current research. Tellurium fiber is used as fiber gain medium,and a modified particle swarm optimization algorithm is proposed,which dynamically adjusts the individual velocity,position and inertial weight values during the iterative process to obtain higher convergence speed and enhance the global search capability.The algorithm is applied to optimize,analyze and verify the configuration of each pump light parameter of Raman fiber amplifier,and finally,the average gain of the back-pumping Raman fiber amplifier is up to 23.738 8 dB and the gain fluctuation is about 0.209 〖ＫＧ－１／６〗8 dB.The results show that the modified particle swarm optimization algorithm can optimize the Raman fiber amplifier pump power and wavelength accurately and quickly,and can improve the performance of Raman fiber amplifier effectively,which hasa certain reference significance for the design of Raman fiber amplifier in the future.

In order to analyze the internal material properties of the isotropic composite material,combined with the theory of laser ultrasound,the simulation software Comsol Multiphysics is used to establish an equivalent model of the transversely isotropic materials excited by the laser thermoelastic ultrasonic wave,by setting a perfect matching layer,low reflection boundary,etc.The method eliminates the boundary echo.The finite element method is used to simulate the physical process of ultrasonic excitation in transversely isotropic materials,and the time-domain analysis of the defect echo signal was carried out.The results show that the increase in the thickness of the plate will gradually transform the ultrasonic wave into a higher-order mode,and as the thickness of the plate increases,the cut-off frequency of the mode will gradually decrease,and the ultrasonic wave propagating in the isotropic composite material will gradually lose the Lamb wave.In turn,it shows the characteristics of surface waves,transforming Lamb waves into surface waves and excites Rayleigh waves.This simulation lays a foundation for further research on the generation of ultrasonic waves and defect location in complex media.