In the future, wireless communication systems will deploy higher frequency bands to meet their development needs, such as high performance, miniaturization, and diversification. Radio frequency (RF) technology constitutes the foundation for the implementation of wireless communication systems and directly determines the performance of the entire communication system. Glass exhibits an excellent high-frequency electrical performance and high stability. As surface high-density forming technology for fine metal wire and high-reliability processing technology for large aspect ratio through hole continue to evolve, glass is gradually finding wide application as a substrate material for advanced packaging of RF systems and passive RF components. This article provides a systematic overview of the latest progress in the research on the applications of glass in three RF packaging technologies: chip back loading, chip embedding, and 3D stacking, as well as five passive RF components: transmission lines, interconnect lines, filters, phase shifters, and antennas. Different advanced packaging structures and passive RF components are analyzed from the perspectives of structural integration, process feasibility, and RF performance, along with the corresponding challenges. Finally, directions for the development of glass-based RF systems in the future is discussed from the aspects of optoelectronic co-packaging, SOP integration, and efficient heat dissipation.

To improve the sensitivity of air pollutants detection, a type of surface plasmon resonance-photonic crystal fiber (SPR-PCF) sensor based on graphene coating is proposed. The sensor adopts a hexagonal double arrangement structure with air holes of different diameters. The outer core of the sensor is coated with gold film, and a layer of graphene material is coated between the gold film and the object to be measured to improve sensitivity. The effects of metal layer thickness, spacing and diameter of air holes, and graphene thickness on the sensitivity of the SPR-PCF sensor were systematically investigated using the finite element method. For a refractive index range of 1.398~1.402, the results show that the sensor exhibits significant surface plasmon resonance effect. Applying the optimal structural parameters obtained by analysis, the sensing performance was almost unaffected by the ambient temperature. The maximum wavelength sensitivity reached 55,000 nm/RIU, and the corresponding refractive index resolution was 1.82×10-6RIU. Compared with the structure coated with graphene in the same refractive index measurement range, the wavelength sensitivity of the proposed SPR-PCF sensor was 2.86~9.17 times that of the existing typical SPR-PCF sensor, and the refractive index resolution of the former was 0.109~0.349 that of the latter. Therefore, the proposed SPR-PCF sensor has good application prospects in the detection of atmospheric pollutant type and concentration.

Insulated gate bipolar transistors (IGBT) are widely used in the electronic industry for their excellent advantages such as low on-state voltage, high current capability, fast switching speed, and high input impedance. The physical model of 3 300 V field cut-off IGBT (FS-IGBT) with planar gate structure was established based on the device structure and physical mechanism presented in this paper. The breakdown performance was simulated, and an in depth analysis was conducted on the transfer and conduction characteristics. Within a reasonable threshold voltage range, the drift region thickness and regional P-body, P+ substrate, and N-type buffer doping concentrations had significant effects on the breakdown voltage and on-voltage drop of the device. The results indicate that the saturation voltage and turn-off loss of the FS-IGBT device were reduced by 5.2% and 32%, respectively, compared with that of the conventional planar-gate IGBT structure, and the breakdown voltage improved by 14%, with a threshold voltage of 5.3 V.

A 8×8 silicon single-photon-avalanche-photodiode (SPAD) module for single photon detection was designed and fabricated. The silicon SPAD focal plane array using the N+-p--P+ structure works in the Geiger avalanche mode, and the pixel pitch was 200 m. The SPAD pixel used the back reflecting mirror and micro-scattering structure to improve the photon detection efficiency at 1 064 nm wavelength. The readout integrated circuit (ROIC) combined the high voltage quenching, time-delayed reset, avalanche current detection, high voltage protection, and other functional circuits on a single chip。The 8×8 silicon SPAD module can detect a single photon. The test results show that the photon detection efficiency of the SPAD module at voltages in excess of 40 V was approximately 11.1% at a wavelength of 1 064 nm; the average dark count rate was 4.6 kHz; and after-pulse probability was 6.76% for a dead time of approximately 100 ns.

A multimode-fiber Mach-Zehnder interference strain sensor with threaded grooves is proposed. The threaded groove structure is made on the multi-mode fiber as the sensing area, and the cross-sectional area of the spiral groove fiber is smaller than that of the single-mode fiber, so that the actual strain on the sensing area is greater than the applied strain. When a quantitative strain is applied to the fiber, the wavelength drift is greater, which increases the strain sensitivity of the sensor. The experimental results show that in the strain range of 0~900 , the strain sensitivities of the single- and multi-mode fiber sensors are -6.44 to -16.89 pm/, respectively, with linearities of 0.979 and 0.981, respectively. The sensor has the advantages of simple structure, high sensitivity, and high mechanical strength, providing a new solution for enhancing the sensitivity of optical fiber interference strain sensors.

Using angle-resolved photoemission spectroscopy (ARPES) combined with a thin-film stretching holder, the temperature-dependent band measurements of bulk black phosphorus were initially conducted, before performing band measurements under tensile strain at 200 ℃. The results show that as the heating temperature increases from 30 ℃ to 200 ℃, the valence band maximum (VBM) gradually shifts towards deeper levels. This change is attributed to the weakening of interlayer interactions due to thermal expansion of the lattice. The strain-ARPES measurements at 200 ℃ indicate that as the tensile strain along the zigzag direction increases, the VBM exhibits a linear shift towards shallower levels, with a shift rate of 17.8 MeV/% strain. This is because at high temperatures, the tensile strain induces a greater lattice contraction in the out-of-plane direction, resulting in a more significant enhancement of interlayer interactions, which subsequently leads to a more pronounced VBM shift.

Zn thin films were deposited on quartz glass substrates using the magnetron sputtering method, whereby ZnS thin films were formed via low-temperature sulfurization annealing. Finally, annealing of ZnS films was conducted at high temperatures of 500~800 ℃ in an argon atmosphere for 1 hour. The effect of annealing temperature on the properties of the ZnS films thus obtained was investigated by X-ray diffraction (XRD), scanning electron microscopy (SEM), energy dispersive X-ray spectrocscopy (EDS), and ultraviolet (UV) visible spectrophotometer. The results show that the ZnS thin film prepared by low-temperature sulfurization has a hexagonal structure. After high-temperature annealing, the grain size on ZnS films increased, and their transmittance in the visible light range was approximately 80% within bandgaps of 3.59~3.63 eV. With further increases in annealing temperature, the grain size of ZnS thin films increased from 20 nm to 28 nm, the S/Zn atomic ratio decreased, and the surface morphology also changed. Sulfur impurities appeared in the ZnS films after high-temperature annealing at temperatures ≥600 ℃. The optimal high-temperature annealing temperature was 500 ℃, and the resulting ZnS film exhibited good film quality.

The third-generation semiconductor material -Ga2O3 is widely used in high-power electronic devices owing to its wide bandgap and excellent Baliga figure of merit. Among many growth methods of -Ga2O3, hydride vapor-phase epitaxy (HVPE) can meet the requirements of growth rate, quality, and cost in preparing large-scale -Ga2O3 wafers. A three-dimensional numerical simulation of the -Ga2O3 growth process in HVPE growth chamber with showerheads was conducted to systematically and effectively evaluate the effect of parameters on growth results. The orthogonal experimental method was introduced to analyze the growth parameters based on computational fluid dynamics (CFD) simulations. The results indicate that growth rate and uniformity are closely related to the substrate tilt angle, O2 inlet rate, N2 inlet rate, GaCl outlet dip angle, and structure of showerheads. This study proposes an optimized combination of parameters, providing a useful reference for achieving a balance between growth rate and growth uniformity of the epitaxial layer in actual experiments.

Aiming at the failure problem of an array detection point of a fibre Bragg grating (FBG) curve shape sensor, this paper constructs a model based on double-array FBGs to solve the curvature and bending direction of the curve shape sensor. Firstly, the function relationship between the centre wavelength drift and curvature is derived, then the mapping relationship between curvature and bending direction is constructed, and finally the Frenet-Serret equation is used to achieve the curve shape reconstruction. On the basis of Ansys modelling and simulation analysis, this paper fabricates the FBG curve shape sensor and analyses the experimental data. The simulation and experimental results verify the feasibility of the dual-array FBG curve shape reconstruction proposed in this paper. Compared with the typical three-array FBG curve reconstruction method, the distal position errors of the simulated arc and s-shape increase only by 0.9% and 0.13%, respectively; and the average distal position error of the experimental data increases only by 0.34%.

Most hyperspectral unmixing methods based on autoencoders mainly focus either on the spatial information or on the spectral information, while ignoring the balance extraction of them. To address this issue, we propose a hyperspectral image unmixing method based on autoencoders and multi-scale spatial-spectral feature encoding. This method utilizes a CNN encoder for multi-scale unmixing feature extraction. The Transformer encoder receives the multi-scale unmixing features. It further utilizes sub-Transformer encoders and a global Transformer encoder to decouple the dependence between spatial and spectral information. Experimental analysis is conducted on two real datasets to validate the performance of the proposed method. The results demonstrate that the proposed unmixing algorithm can improve the accuracy of hyperspectral image unmixing.

To address the issues of missed and false detections in existing dense crowd pose estimation algorithms, an improved YOLOv8sPose algorithm for dense crowd pose estimation, namely, YOLOv8Pose-Dense Crowd (YOLOv8Pose-DC), is proposed. First, a centralized intrinsic adjustment feature pyramid network is designed, which combines deformable attention mechanisms and coordinate attention-based spatial pyramid pooling fast (CASPPF) in a parallel manner, globally focusing and adjusting the pyramid network from top to bottom, thereby increasing the spatial weight of global representation within the network. This enables the improved algorithm to obtain comprehensive and distinctive feature representations. Second, a multi-scale dual detection head structure is proposed, reducing computational complexity while enhancing model detection efficiency. Furthermore, the DySample module is utilized to improve the upsampling efficiency of the model. Finally, a spatial context aware module (SCAM) is added to enhance the model's ability in associating global information and suppressing irrelevant background features, to highlight human characteristics. Compared to the baseline model, YOLOv8Pose-DC increases mAP@0.5 by 3.1% and recall rate by 4.2%. The designed algorithm significantly improves performance and fully meets production requirements.

To address the difficulty in simulations of massive array microwave photonics systems and devices, a distributed cross-domain parallel simulation method is proposed in this paper. First, the pipelined parallel computation across system structural domains is achieved based on the independent transmission between the channels of a massive array microwave photonics system. Second, parallel data computing across time domains is achieved by utilizing the relative independence of pre- and post-processing times. Furthermore, a static load balancing strategy is used to assist in allocating computational resources. These two approaches are effectively combined to achieve highly efficient simulation of microwave photonics systems, which addresses the issue of long simulation times caused by large amounts of data and models. For a 64-array microwave photonics system with more than 400 models, this technique reduces the simulation time from 39 hours to 23 minutes, resulting in a simulation efficiency improvement of two orders of magnitude. This advancement holds potential for significantly reducing the development cycle of microwave photonics engineering prototypes.

Aiming at the problem of low insect identification accuracy and difficult detection of small target insects against the complex background of cotton fields in Xinjiang, a lightweight insect detection model based on YOLOv5s (LID-YOLO) is proposed. First, the GhostNet network is used to replace the original cross-stage partial CSPDarknet53 network in the backbone, and the Slim-Neck module is used to improve the neck network, to achieve a lightweight model. Second, the fusion module BottleNet Transformer is introduced to reduce the number of model parameters and enhance the capability of network feature extraction to better detect small targets. Finally, the normalization-based attention module (NAM) is added to extract detail features by applying a sparse weight penalty to suppress non-significant weights and improve model accuracy. The experimental results show that compared with YOLOv5s, the LID-YOLO model reduces the number of parameters, calculations, and model weight by 30.9%, 45.6%, and 29.7% respectively. The accuracy rate of LID-YOLO model reached 97.4%, and detection speed was 55.25 FPS, which is 1% point and 2.62 FPS higher than that of the original YOLOv5s model. The LID-YOLO model not only ensures lightweight, but also improves detection accuracy to better meet the requirements of crop insect detection.

To improve the signal-to-noise ratio and imaging quality of aerospace TDICCD remote sensing cameras, in-depth research was conducted on the crosstalk between the imaging spectral bands of TDICCD remote sensing cameras. In the design process of high-density TDICCD imaging circuits, owing to the limited space, dense layout, and wiring of PCB circuit boards, the isolation between signals is unsatisfactory, thus resulting in crosstalk between the spectral bands of the signals. This type of crosstalk significantly affects the acquisition of effective signals in various spectral bands. Under different operating frequencies of each spectral band, crosstalk stripes are generated in the image. This article first introduces the composition of an imaging system based on signal processing and describes two types of crosstalk phenomena. Subsequently, the causes of crosstalk and its boundary conditions are analyzed, and a solution for suppressing imaging crosstalk is provided. Finally, imaging and signal-to-noise ratio testing are conducted on an improved TDICCD imaging system. The experimental results show that the measures adopted effectively removed the crosstalk stripe between the TDICCD imaging spectral bands as well as significantly improved the camera signal-to-noise ratio. Under a camera radiance of 133.729 W/(m2×Sr), the camera signal-to-noise ratio increased by 6.92 dB to 47.83 dB, thus satisfying the demands of practical engineering.

A deep fusion generative adversarial network (DF-GAN) enhancement model combined with self-attention mechanism is proposed for low semantic relevance, fuzzy details, and inadequate structural integrity in text-to-image tasks. First, the bidirectional encoder representations from transformers (BERT) model is used to mine the semantic features of text context and combined with the deep text-image fusion block to realize the matching of deep text semantics and image regional features. Second, a self-attention mechanism module is introduced as a supplement to the convolution module at the model architecture level, aiming to enhance the establishment of long-distance and multilevel dependencies. The experimental results demonstrate that the proposed enhancement model not only strengthens the semantic relationship between the text and image but also ensures the inclusion of precise details and the overall integrity of the generated image.

A notable limitation of the local self-similarity descriptor (LSS) is that it is considered unsuitable for multimodal image registration. In order to address this issue, a novel self-similarity descriptor is proposed and effectively applied for multi-modal image registration. First, the phase congruency algorithm is used to extract the maximum moment of the multimodal image. Second, Harris keypoints are extracted from the edge images obtained from the maximum moment information. Third, a binary image is produced based on the edge image, and a binary self-similar descriptor is constructed based on the binary image. Finally, descriptor similarity calculations and keypoint matching are performed for multimodal images. Comparative experiments demonstrate that the proposed binary self-similar descriptor serve as a replacement for the traditional self-similar descriptor, and it effectively improve the compatibility and efficiency of the self-similar descriptor for multimodal images.

To reduce the bias drift of the fiber optic gyroscope arising from the temperature effect and improve accuracy, a temperature compensation model of the fiber optic gyroscope was established based on the radial basis function (RBF) neural network model and particle swarm optimization (PSO-RBF). Temperature compensation tests were conducted on the three-axis fiber gyroscope in temperature environments of -40 to +60 ℃. The experimental results demonstrate that the model reduces the bias drift of the entire process of the fiber optic gyroscope by more than 85% under the condition of variable temperature, with prediction stability and compensation effect better than those of the traditional polynomial and unoptimized RBF models.

Aiming at the issue of high end-to-end latency and the inaccurate of simulation strategy because of the large synchronization delay in 5G power virtual private optical network slicing, a latency optimization algorithm based on deep reinforcement learning is proposed. Firstly, a 5G power virtual private optical network slicing system model is established, which includes a synchronization node for network element updates and other service nodes, where the synchronization node is directly connected to the software defined network controller through a dedicated special fiber. Then, an optimization problem is proposed to minimize the total latency, including business and network element update latency. As this problem involves discrete and continuous variables, both discrete and continuous deep reinforcement learning algorithms were employed for solution. Simulation results show that the proposed algorithm can effectively reduce the latency of the power virtual private optical network slicing network, meet the requirements of service quality, and ensure the real-time performance of the simulation strategy.

The additional phase noise of the microwave optical transmission link seriously deteriorates in strong vibration environments. To address this limitation, firstly, the reasons for the deterioration were initially analyzed. Subsequently, the mechanism of vibration transmission was studied and a math model was established. Simulations and experimental analyses demonstrated that the acceleration sensitivity of the optical link system is on the order of 10-12 in the low frequency range and 10-11 in the high frequency range. The experimental results further show that in a strong vibration environment, the additional phase noise of the optical link system deteriorates to about 40 dBc/Hz and at resonance, to about 48 dBc/Hz. By reducing the acceleration sensitivity of the optical system and avoiding resonance, the effect of vibration on additional phase noise can be effectively reduced.

Detection of concrete bridge surface damage is crucial for bridge maintenance. However, existing machine vision-based methods suffer from low detection efficiency and accuracy when dealing with small-scale damages and complex backgrounds. In this paper, a novel detection network based on YOLOv5 is proposed. By optimizing the YOLOv5 backbone network and introducing a global attention mechanism along with a multiscale pyramid spatial pooling structure, detection accuracy and efficiency are effectively improved, especially in the detection of small-scale damages in complex backgrounds. Experimental results show that average detection accuracy of the improved model increased by 4.1% compared to that of the original network structure, surpassing that of YOLOv7 and YOLOv5 in detection performance on small-scale damages, such as holes and complex backgrounds. The proposed method achieved a 2.3% increase in average detection accuracy and a 30% improvement in detection speed over YOLOv7.

An indoor cooperative localization method for high-density targets based on received signal strength indication (RSSI) is proposed for the indoor environment problem. In indoor environments, the true positions of the assisted localization targets are unknown and the cumulative error of the inertial navigation system becomes large when the localization module of some targets fail. Using the improved interactive multimodel extended Kalman filter (IMM-EKF) algorithm, the auxiliary target position was integrated as an unknown parameter into the state vector of each model. This integration effectively mitigates the effect of the auxiliary target position error on the positioning accuracy and prevents the accumulated error of the inertial navigation system after module failure. Simulation experiments demonstrated that the standard deviation of the positioning error in the X- and Y-directions was reduced by 32.19% and 23.45%, respectively, compared with that of the traditional IMM-EKF cooperative positioning method, thereby improving the positioning accuracy of indoor targets while maintaining a high positioning effect of the localized targets in the case of positioning module failure.

The frequency drift phenomenon exists in both the acousto-optic modulator and laser of the phase sensitive photosensitive time-domain reflection (-OTDR) system, leading to phase shift and amplitude fluctuation in the demodulated signal, which affects the accuracy of vibration signal detection. This paper proposes a -OTDR system structure based on dual acousto-optic modulators, adopting heterodyne detection to construct the local oscillator and sensing signals with the same frequency drift, thereby overcoming the influence of frequency drift in the backward Rayleigh scattering optical signal of the sensing fiber. The optical signal output by the laser is divided into upper and lower branches. The signals in these two branches are then passed through two acousto-optic modulators and two optical splitters, respectively, generating four frequency-shifted optical pulses in total. One of the optical pulses in the lower branch enters the sensing fiber, and its backward Rayleigh scattering signal interferes with one of the optical pulses in the upper branch, generating a sensing beat signal. The other two pulses generate the local oscillator beat frequency signal, which after coherence have the same frequency drift effect. The frequency drift issue is resolved through quadrature phase demodulation. This study establishes an orthogonal demodulation model, to suppress frequency drift. By theoretically deducing and analyzing the mechanism of this model, frequency drift was suppressed. The feasibility of the dual acousto-optic modulator system in suppressing the influence of frequency drift was further verified through simulations.

A multiscale feature fusion-based traffic sign detection (TSD) algorithm is proposed to address the problem of poor performance of existing object detection algorithms in identifying small target traffic signs. First, a novel cascaded multiscale feature fusion network was designed, which fully utilizes the multiscale sequence feature fusion structure and triple feature encoding module, enabling the network to better integrate the detailed and global features of traffic signs. Second, deformable attention mechanisms were incorporated into the backbone network to enable the model to focus on relevant regions and capture richer image features. Finally, the use of the inner intersection over union (IoU) loss function improved the generalization performance of the TSD-YOLO model. The experimental results on the CSUST Chinese traffic sign detection benchmark (CCTSDB) dataset show that the average accuracy of the improved model was 55.3%, which is 3.2% higher than that of YOLOv8s. In addition, the performance on the visual object classes (VOC) dataset and benchmark Tsinghua-Tencent 100K (TT100K) dataset highlights the excellent generalization performance of the model.

This paper presents an integrated multi-parameter sensing system that combines a Brillouin optical time domain reflectometer (BOTDR) and phase sensitive optical time domain reflectometer (-OTDR) to simultaneously measure temperature, strain, and vibration information of the sensing fiber. A double heterodyne detection structure was set up to multiplex the signal, and the scattered light splitting ratio was optimized. For a sensing distance of 8.6 km and spatial resolution of 3 m, the temperature and strain measurement accuracy of the system are 0.48 ℃ and 18.15 , respectively. The vibration measurement results show that the system has a maximum frequency response of 5 kHz and a dynamic strain resolution of 8.48 n/√Hz at 100 Hz.

The extraction of buildings from high-resolution remote sensing images is of great significance in three-dimensional reconstruction of urban scenes. A remote sensing image building extraction shape correction generative adversarial network (SCGAN) with improved high-resolution network (HRNet) is proposed to address the problem of low segmentation accuracy caused by mutual occlusion and blurred boundaries of buildings in complex background remote sensing images when using traditional convolutional methods. Based on the HRNet structure, shape correction units are introduced to enhance the model's perception of building edges and shapes, and adversarial learning strategies are used to strengthen detailed features such as building boundaries and geometric shapes. The experimental results show that the SCGAN model based on adversarial learning and shape correction units effectively improves segmentation accuracy in building extraction, with IoU of 90.94% and 70.89% on the WHU and Massachusetts datasets, respectively, exhibiting the best performance compared to popular semantic segmentation models.

To address the issues of the low detection accuracy, exorbitant number of parameters, and extensive computations in the existing YOLOv8s vehicle detection model, a lightweight car detection-YOLO (LCD-YOLO) algorithm based on improved YOLOv8s is proposed for lightweight vehicle target detection. The algorithm applies frequency-adaptive dilated convolution (FADC) to optimize the cross-range partial (CSP) bottleneck with two convolutions (C2f) in YOLOv8s to enhance feature fusion ability. Shared convolutional layers reduce the number of network convolutions and parameters, thereby achieving a lightweight model. Through the dynamic focusing of the bounding box regression loss calculation method, this model can effectively improve the network's ability to detect occluded overlapping targets and improve the accuracy of border detection. Experiments on the KITTI dataset show that the average detection accuracy of the proposed algorithm is improved to 95.1%, which is 2.9% higher than that of the YOLOv8s algorithm, while reducing the number of network parameters by 14.9% and amount of computation by 10.9%, which can better satisfy the actual detection needs of vehicles.