Abstract：Quantum computing is a cutting-edge topic that has attracted much attention in the current computing field. Quantum cutting is a highly promising computing framework proposed to break through the current hardware limitations and increase the number of available quantum bits. In the research and verification of it using classical computers, quantum simulation of subcircuits is the most important bottleneck link in computing power. In order to improve the efficiency of quantum simulation, this study found that the naive linear algebra method used in the key matrix calculation link of Qiskit Aer, a quantum simulation framework commonly used in quantum projects, has room for optimization using Single Instruction Multiple Data (SIMD). Based on this, this paper uses AVX2 instructions to optimize the code of the matrix calculation part, and tests the control variables through the containerization method. The SIMD optimization method successfully gave a significant improvement in the quantum simulation efficiency of Qiskit Aer. It has been verified that this improvement is stable, reliable, and reproducible, and will not introduce unknown effects on irrelevant links outside the quantum simulation. The research results of this paper improve the efficiency of quantum cutting simulation, provide a faster tool for the study of quantum cutting, and provide experience and paradigms that can be used for further optimization of the Qiskit framework.

Abstract：Addressing the challenge of low development efficiency in the heterogeneous resource architecture combining general-purpose CPUs and Field Programmable Gate Arrays (FPGAs), a novel proxy-based thread management framework is proposed by leveraging existing software and hardware thread parallel computing paradigms based on the Linux operating system kernel, enabling the unified management of hardware and software threads. This innovation leads to a parallel computing system architecture that can realize the flexible reconfiguration of FPGA resources using. The proposed architecture ensures functional thread isolation and interface separation, supporting concurrent development of multiple functional threads. This article verifies the validation of the system by conducting the hardware thread redesign of Strassen's matrix multiplication algorithm and bubble sort algorithm, which demonstrates the system's ability to achieve functional thread decoupling and significantly improve the development efficiency of system software functionalities.

Abstract：This paper focuses on the issue of limited multi-core parallel scalability caused by thread lock contention mechanisms in the architecture of the Ceph storage system's Object Storage Device (OSD). It conducts research on collaborative optimization technologies between the next-generation Crimson-OSD architecture and SmartNICs, proposing a hierarchical cooperative optimization framework. Related studies demonstrate that employing SmartNIC-based cooperative optimization achieves a 70% reduction in CPU utilization through RDMA network offloading, while heterogeneous computing engines enable hardware acceleration for erasure coding, improving data recovery speed by 4.84 times. The research outcomes provide theoretical foundations and key technical references for hardware acceleration in distributed storage systems, offering guidance for optimizing storage systems in data-intensive scenarios such as high-performance computing and cloud-edge-end integration.

Abstract：In response to the complex and changeable marine environment, such as fog, strong light reflection, low illumination at night, and the limited characteristic information of small targets at sea, a small target detection method based on YOLOv11 for marine multispectral features is proposed. By designing a dual-branch YOLOv11 model to handle the fusion of cross-modal data features and introducing a global attention mechanism module for training, the improved multispectral image small target detection model can fully utilize the multispectral features, achieving accurate detection and positioning of small target objects in multispectral images, especially performing well from the perspective of unmanned aerial vehicle (UAV) aerial photography. This cross-modal fusion method can significantly enhance the robustness and accuracy of small target detection at sea. Experiments show that the improved model can achieve an mAP@50 of 96.5% on the VTSaR dataset, an increase of 0.4% compared to YOLOv11n, providing a new solution for the detection of small targets in marine aerial unmanned search and rescue.

Abstract：Based on the actual needs of landing support in UAV after-sales service, this paper carries out research on technologies that use solar-blind ultraviolet detectors to support UAV after-sales maintenance and support. Through the analysis of atmospheric transmission characteristics in the ultraviolet band, the design of ultraviolet detection system, the design of ultraviolet cooperative beacon points, and the experimental verification of P4P photogrammetry algorithm, it provides technical support for landing performance evaluation, fault diagnosis, and maintenance effect verification in UAV after-sales support, and verifies the feasibility and rationality of this technology in ensuring the reliability of UAV landing in after-sales service.

Abstract：In response to the problems existing in the detection of maritime crashed aircraft targets, such as insufficient background noise suppression, incomplete target contours, weakened target features, and difficulty in identifying small target pixels, a dual-stream fusion gray value discrimination algorithm based on time difference accumulation optimization is proposed. By constructing a three-dimensional optimization of "dual-target feature enhancement-noise suppression-spatiotemporal registration", a five-level progressive image processing is realized. An airborne platform image acquisition and evaluation system under typical scenarios is built; "day-night" stepwise experiments are conducted, and it is found that the maximum dual-target signal ratio in daytime experiments can reach 1.739, and the signal-to-background ratio in nighttime environments can reach 25.

Abstract：A multifunctional clock fan out buffer was designed based on CMOS technology. The buffer is equipped with a programmable frequency divider and delay adjuster, which can independently output differential clocks in 4 channels. Each channel can be divided and delayed for adjustment, and all support three logic level types: LVDS (MAX 1.75 GHz), HSTL (MAX 1.75 GHz), and 1.8 V CMOS (MAX 350 MHz). The testing results show-that: 1.75 GHz differential clock input/output; each output can bypass the frequency divider or set an integer division ratio of up to 2 048; each channel can be adjusted for both digital and analog delay; broadband random jitter<110 fs RMS; additional random jitter of 39 fs RMS (typical value, 12 kHz~20 MHz). It can meet the low jitter requirements for applications such as data converters and clock trees, and can be widely used in wireless transceivers and communication systems.

Abstract：A compact and ultra-wideband 5~18 GHz multifunction chip in Gallium Arsenide (GaAs) technology is presented. The multifunction chip integrates 3 Single Pole Double Throw (SPDT) switches, 3 medium amplifiers, 6 bit digital attenuator and 6 bit digital phase shifter. All pass filter and high/low pass filter topology is adopted in phase shifter design to improve the phase setting accuracy and reduce the chip size. Test results show that multifunction chip has high RF performance and is suitable for applications in phased array radar. The chip size is 5.3 mm×3.5 mm×0.1 mm. The currents consumption is 70 mA at +5 V supply voltage and 5 mA at -5 V supply voltage.

Abstract：In order to meet the application requirements of JESD204B/C system in high-speed multi-channel scenarios, a programmable frequency divider with high speed and wide operating frequency is designed by using current mode logic (CML). Based on the 0.18 μm SiGe BiCMOS process, the results of the chip test show that the frequency division can be achieved 2~16 380 times in the input frequency range of 300 MHz~16 GHz, and there are three different modes of output: signal generator, pulse generator and repeater.

Abstract：Aiming at the problem of significant nonlinear error of traditional fluxgate sensor in wide range current measurement, a collaborative correction method based on field programmable gate array (FPGA) high-precision time difference detection and polynomial dynamic compensation is proposed. In this study, by establishing the mapping relationship between the positive and negative saturation time difference of the magnetic core and the measured current, a digital FPGA processing architecture is constructed to capture the saturation time difference in real time, and a polynomial compensation model including nonlinear effects is established. The experimental results show that this sensor can precisely detect complex and weak leakage currents. The determination coefficient R2 of the compensation model is 0.999 976, which is 0.11 percentage points higher than that of the linear model. The root mean square error is reduced by 85.4 %. Through hardware-algorithm collaborative optimization, the accuracy drift in industrial field environment is effectively suppressed, which provides a high-precision and low-cost solution for device-level current monitoring of smart grid.

Abstract：In the application of Kalman filtering algorithm for real-time processing of measurement data, methods are often employed to adjust the filter gain matrix in order to address divergence issues. In real-time data processing, it is not possible to determine the gain coefficients of the filter gain matrix adjustment through a posteriori means; therefore, an adaptive determination method targeting the data must be designed. This paper examines the error characteristics of the data sequence, adjusts the filter memory decay step size, determines the filter memory decay coefficient, and employs the hyperbolic tangent (tanh) function to calculate the gain coefficients. Simulation results demonstrate that the Kalman filtering algorithm with adaptive gain coefficients can effectively adapt to common measurement data and is suitable for real-time processing of measurement data.

Abstract：According to the strategic requirements of the country for new quality productivity of enterprises, and with the increasing development of power grid intelligence, the demand for dual-mode modules equipped with low-power wireless communication channels and carrier communication channels is gradually increasing. Traditional production tools suffer from low yield rates and low levels of intelligence, leading to issues such as low production capacity, high costs, cumbersome processes, and high rework rates in the production and testing of dual-mode modules. This article elaborates on the design and implementation of a dual-mode module intelligent production testing system platform for power and IoT, based on actual research and development projects. ...

Abstract：Current intrusion detection methods suffer from challenges such as non-linear correlations in high-dimensional data and severe class imbalance, resulting in high missed detection rates for rare attacks and poor generalization. To address these issues, this study proposes a hybrid approach combining generative adversarial networks (GAN) and ensemble learning. GAN generates synthetic attack samples aligned with real data distributions to mitigate imbalance, while a sparse autoencoder (SAE) reduces feature dimensions non-linearly, enhancing discriminative power. A Bagging ensemble framework integrates K-means, hierarchical clustering, and Gaussian mixture models (GMM), with DBSCAN meta-learning refining clustering outputs for robustness. Evaluations on the KDD CUP99 dataset demonstrate significant improvements: minority-class representation rises from 2.6% to 17.7%, achieving 96.4% accuracy, 7.3% false detection rate, and over 97.6% recall for critical attacks (e.g., U2R, R2L). This method outperforms traditional techniques, offering a novel solution for securing modern power systems against evolving cyber threats.

Abstract：This paper presents a dual-beam reflectarray antenna operating at 10 GHz, which is designed using hybrid resonant elements combining Archimedean spirals and sinusoidal structures. Based on the principles of coding metasurfaces, a 16 × 12 array employing the encoding sequence “0110…” is developed. A Vivaldi antenna is adopted as the feed source, achieving stable dual-beam radiation characteristics. A prototype antenna was fabricated and measured. Experimental results indicate that the proposed antenna exhibits a -10 dB impedance bandwidth of 7.5% (9.56~10.31 GHz) with excellent impedance matching. At the operating frequency, the antenna demonstrates symmetric dual-beam radiation patterns in the E-plane with stable gain performance. Specifically, the measured 3-dB beamwidth is 9.7°, the dual-beam deflection angle is 13.5°, and the peak radiation gain reaches 13.0 dBi. The use of a hybrid harmonic oscillator unit solved the problem of limited bandwidth in traditional reflect array phased arrays. The application of coding beyond the surface further simplified the complexities of beam control.

Abstract：With the rapid development of 6G communication technology, the space-air-ground integrated network has become a research hotspot due to its capability of global seamless coverage. This paper systematically investigates the development trend of coding technologies in the 6G space-air-ground integrated network based on patent analysis, focusing on four major technical branches: Turbo codes, LDPC codes, Polar codes, and AI coding. By analyzing global patent data from 2016 to 2025, the paper reveals the trends in technological evolution, regional competitive landscape, and distribution of innovative entities. The findings show that South Korea leads in the number of patents, China has an advantage in Polar codes, and AI coding accounts for more than 50% of the patents but is still in the early stages of standardization. The paper also identifies key technological gaps such as high dynamic channel adaptation and space-ground collaborative optimization, providing strategic references for the development of 6G coding technologies and standard setting.

Abstract：Object detection has broad application prospects in the field of remote sensing. Although object detection algorithms have made significant progress in natural images, these methods still face numerous challenges when directly applied to remote sensing images. The background of remote sensing images is often complex, and the objects are relatively small, which leads to an extremely imbalanced distribution of foreground and background information. To address the issues of small targets and object occlusion in drone images, this paper proposes an improved drone small object detection algorithm based on PinwheelConv. To enhance the model's performance in detecting small objects, the PinwheelConv is used in place of regular convolution in the backbone network, which better adapts to the extraction of small target features. Additionally, a C2f-PC module based on the windmill convolution idea is designed to replace the C3k2 module in the backbone. To address the severe occlusion problem in drone images, this paper innovatively introduces the C2f-PDWR module to replace the C3k2 module in the neck network, enhancing the model's feature fusion capability. Moreover, a Spatially Enhanced Attention Module (SEAM) is incorporated to improve the model's detection of occluded objects. Finally, this paper proposes a more efficient small object detection model, YOLO-PDS, based on YOLOv11. The proposed method improves the mAP50 by over 3.7% and the recall rate by more than 2.2% compared to the baseline YOLOv11 detection method on the VisDrone2019 dataset.

Abstract：Data visualization plays an important role in meteorological big data. Through the display of data panel, users can visually and conveniently understand the basic situation of complex data. However, due to the variability of the location and size of different data sections of the panel, it is difficult for users to quickly find the location of various data sections. Based on this, this study proposes a real-time detection method of data visualization area based on improved MobileNetV3-SSD.On the basis of standard SSD, the attention mechanism is combined, and the improved FPN module is used to fuse the feature information, which can efficiently locate the “Title” position of the target data section, and then display that area for users. The proposed model can significantly reduce the number of parameters, and the precision ,recall and mAP indexes reach 83.05%, 85.02%, 74.35% respectively, meanwhile, the inference time based on CPU is reduced to 244 ms per image.

Abstract：Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that affects social interaction, communication, and behavior. Early diagnosis is crucial, but challenges remain in handling multimodal brain data. To address this, this study proposes an end-to-end model, SABNet, based on multimodal feature fusion. The model uses sparse autoencoders for feature selection on functional data and combines structural data to construct joint features. Finally, bidirectional Long Short-Term Memory (LSTM) networks and attention mechanisms are employed to extract important information from dynamic sequences for efficient classification. Experiments on a medium-sized ASD-MRI dataset show that SABNet outperforms traditional methods in key metrics such as classification accuracy (91.21%). Principal component analysis further validates its effectiveness. This study demonstrates the excellent performance of SABNet in ASD recognition and highlights the potential of multimodal feature fusion in brain disease classification. Future research will focus on optimizing the model architecture, expanding the dataset, and improving its applicability and generalization ability.

Abstract：Deep neural networks have been widely used in photovoltaic power prediction, but they are vulnerable to adversarial attacks. In order to improve the robustness of the prediction model, an adversarial attack algorithm based on fast gradient sign method and a defense algorithm based on adversarial training are proposed. The fast gradient sign method generates adversarial samples with temporal correlation, and establishes the quantitative relationship between attack intensity and prediction error. Adversarial training enhances the generalization ability of the model to input disturbances by combining adversarial samples to resist adversarial attacks. Experimental data show that adversarial attacks can significantly reduce the prediction accuracy of the model, and the model trained by adversarial training can effectively improve the robustness. This method verifies the effectiveness of the countermeasure attack and defense mechanism in photovoltaic power prediction, and has practical application value for the safe operation of power system.

Abstract：Previous CO2 detection systems were large in size, inconvenient to integrate, generally unreliable, and poor in accuracy, making it difficult to meet the detection requirements in the consumer electronics and automotive fields. This paper designs a high-precision and high-reliability carbon dioxide gas sensor module with temperature compensation function, with infrared light source, gas chamber, dual-channel thermopile detector and single-chip microcomputer as the core. In terms of structure, a reflective gas chamber structure is adopted to improve the integration of the system and realize the miniaturization of the sensor. In terms of hardware design, a variety of anti-interference measures are taken, such as magnetic beads, capacitors, shielding layers, etc., to improve the stability of the sensor module. In terms of software, the ratio calibration method is used to establish a temperature compensation model to eliminate the interference of ambient temperature. The experimental results show that: the sensor has a maximum detection error of less than 3% and a response time of less than two minutes in the concentration range of 0~5 000 ppm at a temperature of -20 ℃~50 ℃. It has the advantages of high precision, fast response, miniaturization, and good long-term stability.

Abstract：To accurately measure the depth information of coal bunker level and avoid the impact of high or low level on coal transportation process, a coal bunker level depth detection system based on dual axis laser scanning is proposed. The system adopts a dual axis galvanometer structure to achieve wide field scanning and complete laser ranging at any position on the object surface. We have built a synchronous optical path structure that conforms to the dual axis galvanometer and laser ranging, and designed a focusing module for beam shaping. In the dynamic focusing test, the variation of the light spot within the range of 2 m~5 m was tested, and the diameter of the light spot was all less than 0.5 mm, which meets the design requirements for focusing and illuminating the test object surface. In the depth experiment of the object surface, the height of the object surface was tested at distances of 2 m and 5 m, and the measured deflection angle was between 0°~ 20°. The experimental results show that under two testing conditions, the maximum error is 23 mm, and the error of the test results for the upper and lower limits of the material level is less than 0.5%. The dual axis laser scanning system has high accuracy, a large field of view, and the ability for online real-time testing, which has high practical value in the field of coal bunker level detection.