Reconfigurable intelligent surface (RIS) is one of the potential key technologies for sixth generation (6G) communications, which has the characteristics of low cost, low complexity, and easy deployment. By applying control signals to adjustable elements on the electromagnetic unit, it has the ability to adjusting the wireless communication environments, which provides a new opportunity to improve the high energy-efficiency performance of wireless communication systems. This paper provides a comprehensive overview of channel modeling and characteristics analysis for RIS-assisted unmanned aerial vehicle (UAV) high energy-efficiency communications. Firstly, based on the research basis of the UAV communications technologies, we clarify the necessities of introducing the RIS into UAV communications. Then, we summarize the key technologies for channel modeling and characteristics analysis for RIS-assisted UAV communications. Finally, we point out some future research directions in RIS-assisted UAV channel modeling and characteristics analysis.

Approximately 71% of the Earth’s surface is encompassed by aqueous elements, such as rivers, lakes, and seas. Concurrently, terrestrial imaging contends with the influence of water in the forms of clouds, snow, rain, and fog. Notwithstanding, contemporary machine vision research and application systems predominantly concentrate on visual tasks within aerial and vacuum environments, leaving a dearth of systematic investigation into visual tasks within various aquatic contexts. Water-related vision, emblematic of water-based optical technology in the realm of vision, is committed to dissecting the scientific intricacies of light-water interactions and their inter-medium propagation. It also entails intelligent processing and analysis of visual image signals within aquatic settings. This discipline concurrently addresses engineering and technical intricacies intrinsic to the progression of advanced, intelligent water-related vision apparatus. Embarking from the fundamentally significant scientific query, “What is the reason for the ocean’s blue color?” this paper proffers an exhaustive survey encapsulating the repercussions of seawater’s light absorption, scattering, and attenuation mechanisms upon underwater visual tasks. Furthermore, the current methodologies for the processing and refinement of subaquatic images are systematically examined. Exploiting the optical attributes of water and factors contributing to image degradation, this manuscript underscores our team’s milestones in pioneering indispensable technologies for underwater imaging and image analysis. Substantial headway has been achieved in devising underwater observation and analytical apparatus, encompassing the full-ocean-depth ultra-high-definition camera “Haitong,” the full-ocean-depth 3D camera, and the full-ocean-depth high-definition video camera. These innovations have distinctly established a comprehensive and methodical proficiency in optical detection within submerged contexts, encompassing variables of color, intensity, polarization, and spectral analysis. This collective endeavor effectively bridges the gap in China’s full-ocean-depth optical detection technology, propelling the progress of exploration and technological innovation within the domain of water-related vision, which offers remarkable application value and societal advantages.

Metamaterials, having important theoretical research and application value in terahertz band, are artificial electromagnetic materials with special properties, which can regulate the frequency, amplitude, phase and polarization of electromagnetic waves. Due to the complexity of design process of metamaterials and the limitation of simulation time, the design of metamaterials is always involved in great challenges. In view of the fact that terahertz metamaterial devices have made certain achievements in biomedicine, broadband communication, security screening and other fields, this paper firstly describes the research progress and problems existing in the design process of terahertz metamaterial devices in traditional design methods, and sums up in detail the research results of encoded metamaterials, especially encoded hypersurfaces and programmable hypersurfaces. Additionally, the application of deep learning algorithms in THz (TeraHertz) metamaterial structure design is summarized. Finally, the challenges and expanding research directions of intelligent method in THz metamaterial structure design are discussed. This research not only provides a reference for people to fully grasp the simple, fast and intelligent design methods, but also puts forward some novel ideas for the development and application of intelligent design methods in terahertz metamaterials.

The terahertz band lies between the microwave-millimeter wave band and the optical band, and its related technologies are of great significance for the development of next-generation high-speed communication, high-resolution imaging, intelligent sensor-communication integration systems and other information fields. The development of terahertz application technology requires terahertz beam manipulation and information loading. As a new realization scheme and device, active metasurface is one of the important ways to realize these functions. In this paper, the terahertz dynamic metasurface is classified according to the regulation function of the metasurface, the amplitude regulation, phase regulation, polarization regulation, beam regulation and high-order nonlinear regulation of the terahertz dynamic metasurface are introduced from the complementary materials technology, structure setup, and working mechanisms and their performance in the application demonstrations. The similarity and uniqueness of different types of dynamic metasurface are introduced, and some of the performances are compared. At the end of this paper, the development trend of the terahertz dynamic metasurface is forecasted. It is hoped that this paper helps more scholars get knowledge of terahertz dynamic metasurface hence promoting the development of this field.

Based on the theory of Jones matrix and the principle of Pancharatnam-Berry phase, a dual-band terahertz circularly-polarized (CP) absorbing and anomalous reflecting chiral metasurface is proposed in this paper. The CP chiral absorption and conversion function with anomalous reflection angles are realized independently in the two frequency bands. The metasurface unit consists of a combination of two chiral structures, which has an absorption rate of 96.3% for the incident left-hand circularly polarized (LHCP) waves, and realizes co-polarized reflection for the incident right-handed circularly polarized (RHCP) waves at a low frequency of 2.53 THz; meanwhile, the co-polarized reflection will be achieved for the LHCP waves at a high frequency of 3.43 THz, and the absorption rate of the RHCP waves is 90.9%. According to the principle of Pancharatnam-Berry phase, the full coverage of the 360° reflection phase can be realized by rotating the chiral metasurface unit. This chiral metasurface array can absorb the incident CP waves with specific handedness at two operating frequencies (low frequency 2.53 THz and high frequency 3.43 THz), and generate the handedness-preserving anomalous reflections with -26° and +19° angles for the corresponding orthogonal CP waves, respectively. This terahertz multi-function beam control device based on the chiral metasurface has numerous potential in electromagnetic energy harvesting, polarization converters, chiral sensing, radar and other fields.

In this paper, the recent research progresses on terahertz metasurface lenses (THz ML) based on artificial microstructure are introduced. The interaction between artificial microstructure unit and THz wave is analyzed by numerical simulations. And the three main existing phase modulation mechanisms, including resonance phase, geometric phase and transmission phase, are described. From the spatially phase distribution of lens, the manipulation of incident THz wavefront can be realized by the suitable artificial microstructure layout, so as to realize the focusing and imaging of THz ML. Due to the advantages of flexible design, ultra-thin thickness and multi-functions, THz ML will have great potential applications in the fields of nondestructive detection, high-speed wireless communication and public security. Considering the key technical parameter of focusing efficiency, the research works of THz ML based on single-layer structure, multi-layer structure, all dielectric structure and tunable materials are introduced, and the trends of future development of THz ML are prospected.

A terahertz polarizer based on bilayer metal subwavelength-grating is designed to achieve high transmission and good extinction ratio. The polarizer is fabricated on the upper and lower surfaces of thin quartz substrate by micro-fabrication technique. Both experimental and simulation results show that the bilayer metal grating has similar transmittance, higher degree of polarization and extinction ratio than the single-layer metal grating. Measured transmittance is between 83.4% and 62.7% in the frequency range of 0.3~2.0 THz, and the degree of polarization and extinction ratio is more than 99.7% and more than 29 dB, respectively. In addition, the two designed bilayer metal gratings are successfully used in the terahertz time-domain spectroscopy (THz-TDS) system, and the degree of polarization of more than 96.2% and the extinction ratio of more than 17.1 dB are obtained. The transmittance, degree of polarization and extinction ratio can be improved by further tuning the parameters (such as the pitch, line width, and metal film thickness) of the metal subwavelength-grating.

Small object detection is an extremely challenging task in computer vision. It is widely used in remote sensing, intelligent transportation, national defense and military, daily life and other fields. Compared to other visual tasks such as image segmentation, action recognition, object tracking, generic object detection, image classification, video caption and human pose estimation, the research progress of small object detection is relatively slow. We believe that the constraints mainly include two aspects: the intrinsic difficulty of learning small object features and the scarcity of small object detection benchmarks. In particular, the scarcity of small object detection benchmarks can be considered from two aspects: the scarcity of small object detection datasets and the difficulty of establishing evaluation metrics for small object detection. To gain a deeper understanding of small object detection, this article conducts a brand-new and thorough investigation on small object detection benchmarks based on deep learning for the first time. The existing 35 small object detection datasets are introduced from 7 different application scenarios, such as remote sensing images, traffic sign and traffic light detection, pedestrian detection, face detection, synthetic aperture radar images and infrared images, daily life and others. Meanwhile, comprehensively summarize the definition of small objects from both relative scale and absolute scale. For the absolute scale, it mainly includes 3 categories: the width or height of the object bounding box, the product of the width and height of the object bounding box, and the square root of the area of the object bounding box. The focus is on exploring the evaluation metrics of small object detection in detail from 3 aspects: based on IoU (Intersection over Union) and its variants, based on average precision and its variants, and other evaluation metrics. In addition, in-depth analysis and comparison of the performance of some representative small object detection algorithms under typical evaluation metrics are conducted on 6 datasets. These categories of typical evaluation metrics can be further subdivided, including the evaluation metric plus the definition of objects, the evaluation metric plus single object category. More concretely, the evaluation metrics plus the definition of objects can be divided into 4 categories: average precision plus the definition of objects, miss rate plus the definition of objects, DoR-AP-SM (Degree of Reduction in Average Precision between Small objects and Medium objects) and DoR-AP-SL (Degree of Reduction in Average Precision between Small objects and Large objects). For the evaluation metrics plus single object category, it mainly includes 2 types: average precision plus single object category, OLRP (Optimal Localization Recall Precision) plus single object category. These representative small object detection methods mainly include anchor mechanism, scale-aware and fusion, context information, super-resolution technique and other improvement ideas. Finally, we point out the possible trends in the future from 6 aspects: a new benchmark for small object detection, a unified definition of small objects, a new framework for small object detection, multi-modal small object detection algorithms, rotating small object detection, and high precision and real time small object detection. We hope that this paper could provide a timely and comprehensive review of the research progress of small object detection benchmarks based on deep learning, and inspire relevant researchers to further promote the development of this field.

With the advent of the big data era and the increasingly in-depth study of the human brain system, the field of neuromorphic computing has made breakthrough progress, offering hope to break through the performance issue of traditional computers at the root level. Neural synapses are important basic units for memory training and data processing in the human brain, therefore, it is of great significance for research on neural morphological devices and the implementation of neuromorphic hardware design to develop new materials and structures to study the plasticity of neural synapses based on novel artificial materials and optoelectronic devices. This paper firstly points out the main performance issue of Von Neumann architecture, draws forth the concept of brain-like computing, puts forward the main performance advantages of neuromorphic devices, and sorts out the development history of neuromorphic devices. Then in the field of memristors, the types of memristors, memristor structures and memristor mechanisms are described and analyzed, the advantages and disadvantages of several types of memristors are compared, and the examples of applications of memristors in different fields are presented. Next, based on neural morphological devices, the structures, working principles and applications of magnetic tunnel junctions, new floating gate transistors, and ferroelectric transistors are selected to introduce. Finally, this paper summarizes the achievements and directions of the current development of neural morphological devices and predicts the development prospects of the industry.

Von Neumann computer architecture faces the bottleneck of “storage wall”, which hindering the performance improvement of AI (Artificial Intelligence) computing. Computing-In-Memory (CIM) breaks the limitation of “storage wall” and greatly improves the performance of AI computing. At present, CIM schemes have been implemented in a variety of storage media. According to the type of calculation signal, CIM scheme can be divided into digital CIM and analog CIM scheme. CIM has greatly improved the performance of AI computing, but the further development still faces major challenges. This article provides a detailed comparative analysis of CIM schemes in different signal domains, pointing out the main advantages and disadvantages of each scheme, and also pointing out the challenges faced by CIM. We believe that with the cross level collaborative research and development of process integration, devices, circuits, architecture, and software toolchains, CIM will provide more powerful and efficient computing power for AI computing at the edge and cloud ends.

In this paper, the packaging technique of terahertz monolithic integrated power amplifier is studied, i.e. waveguide-to-microstrip vertical transition and mode-resonance suppression techniques. Different from the conventional horizontal transition based on rectangular probes, a vertical transition based on a bifurcated probe is proposed, which is suitable for multilayer circuits. To verify the performance, a back-to-back transition is fabricated and measured. In the range of 170~260 GHz, the measured return loss is better than 16 dB, while the single insertion loss, including the loss caused by non-ideal metallic contact, is around 0.42 dB. To further reduce the transition loss, a resonance ring with a slot is proposed to avoid the electromagnetic leakage. As a result, the simulated insertion loss is reduced by half. Moreover, the issue of mode resonance in the amplifier cavity is studied, and electromagnetic band gap structures are set above the plane transmission line to suppress the excitation, transmission and resonance of higher modes. The above techniques are applied to a power amplifier which is operated at 210~230 GHz. In the measurement, the maximum small-signal gain of 20.75 dB and the single packaging loss of 0.8 dB are observed at 210 GHz. At 217 GHz, the maximum output power of higher than 15.6 dBm is achieved, which is consistent with the manual.

In this paper, a robust secure beamforming scheme based on imperfect channel state information (CSI) is proposed for the intelligent reflecting surface (IRS)-aided terahertz satellite communication systems, in order to improve the physical layer security performance of the system. First, we consider a scenario where an IRS-aided terahertz satellite system adopts multicast technique to serve multiple legitimate users in the presence of multiple eavesdroppers in the coverage area, based on which a joint optimization problem is formulated in order to minimize the satellite transmit power under the constraints constructed by the achievable rate (AR) and the achievable secrecy rate (ASR) requirements of each legitimate users. Secondly, considering that AR and ASR are non-convex probabilistic constraints caused by imperfect CSI, an approach associated with the second-order Taylor expansion and S-procedure is proposed to transform the non-convex constraints. Further, the joint optimization scheme of the phase shift of IRS and the transmit power of satellite can be solved by semidefinite programming. Finally, the effectiveness and superiority of the proposed scheme are verified by simulation results.

In this paper, a novel SSPP (Spoof Surface Plasmon Polariton) filter is proposed, fabricated and measured based on 0.18 μm CMOS technology. Meanwhile an extra terahertz SSPP filter is designed and its full-wave simulations are provided to prove feasibility in THz band. This novel SSPP filter has a passband of 11~12.3 GHz (S ₁₁<-10 dB, S ₂₁>-3.5 dB), and its size is compact with only 0.018 4 λ _g × 0.008 4 λ _g in electrical size, making it much smaller than other passive filters designed in IC technology. A designed terahertz frequency band SSPP filter, with a full EM simulated passband of 210.8~241.3 GHz (S ₁₁<-10 dB, S ₂₁>-4.7 dB), possesses an in-band insertion loss less than 2.7 dB and an superior out-of-band suppression. Both types of SSPP filters adopt a new energy transfer method of non-contact electromagnetic coupling, with a novel structural design. Moreover, the miniaturization advantage of the microwave section SSPP coupled filter is obvious, with an electrical size of only 0.019 λ _g × 0.009 λ _g, easy to integrate with chips. Equally scaled novel filters proposed in this work can work in microwave, millimeter wave and terahertz frequency band, which could be a reference for the research of on-chip filters in the future.

In the microwave frequency band, switches applied to reconfigurable metasurface units often have good on-off performance, and can achieve 1-bit phase quantization characteristics with low loss and 180° phase difference. As the frequency increases to the millimeter wave or terahertz band, the switching performance will degrade due to parasitic effects, which makes the design of reconfigurable metasurface units difficult. In this work, the equivalent model of the reconfigurable metasurface unit is a microwave two-port network. By analyzing the impedance parameters and scattering parameters of the network, the transfer relationship between the unit reflection coefficient and the switching reflection coefficient is pointed out. The proposed relationship is verified by simulation of different switches and unit structures in different frequency bands. This work provides the upper limit of the performance of a 1-bit unit according to arbitrary switching parameters. The proposed design procedure can guide the design and optimization of terahertz reconfigurable metasurface element.

In this paper, a reconfigurable terahertz ultra-wideband absorber with umbrella structure based on graphene and vanadium dioxide is proposed, which greatly expands the absorption bandwidth of the terahertz absorber and has the advantages of functional reconfiguration and large modulation depth. The absorber consists of an umbrella shaped VO₂ patch, a polyethylene cycloolefin copolymer (Topas) dielectric layer, graphene layer, and a metallic reflector. When VO₂ is in the insulating state and the Fermi level of graphene is 0 eV, the absorber exhibits total reflection characteristics in the whole terahertz band. When VO₂ is in the metal state and the Fermi level of graphene is 0.75 eV, the absorber achieves ultra-wideband absorption in the frequency range of 3.57~10 THz, with an average absorption rate of more than 94% and an absorption bandwidth of up to 6.43 THz. By adjusting the conductivity of VO₂ and the Fermi level of graphene at the same time, the maximum modulation depth of 97.9% (at 4.3 THz) and 96.8% (at 8.25 THz) can be achieved at the two perfect absorption points, which has good switching characteristics. In addition, the absorber has polarization insensitivity and wide-angle absorption characteristics. In the range of 0~35°, the absorptivity of the absorber is more than 90% in the range of 3.57~10 THz bandwidth. The ultra-wideband reconfigurable terahertz absorber will have potential applications in tunable broadband absorber, stealth devices, thermal detection, terahertz switches and other fields.

This report provides a comprehensive summary and analysis of the application and funding statistics in various programs within the field of information acquisition and processing under the grant application code of the Information Science Department of the National Natural Science Foundation of China in 2023. These programs include the general program, young scientists fund, fund for less developed regions, key program, excellent young scientists fund, and national science fund for distinguished young scholars. Additionally, the report outlines a detailed work plan for the upcoming year's fund, offering strategic recommendations and insights for further development in the field.

As one of the important research directions in the field of computer vision, image classification has a wide range of applications. The success of deep learning-based image classification techniques depends on a large amount of annotated data. However, the cost of data annotation is often expensive. Active learning is a machine learning method that aims to achieve the expected model performance with as few high-quality annotated data as possible, and it can alleviate the problem of high annotation costs and difficulty in obtaining a large amount of annotation information in supervised learning tasks. Based on a sample selection strategy, active learning for image classification selects samples from the unlabeled dataset which are informative and thus contribute more to the training of the classification model, in order to update the annotated training data pool. This process is repeated until a given stopping condition is met or the model annotation budget is exhausted. This paper provides a comprehensive survey of the active learning image classification algorithms published in recent years. According the strategies applied in sample data processing and model structure optimization, existing algorithms are classified into three categories: algorithms based on data augmentation, including those using image augmentation to expand the scale of training data or using the differences in image feature interpolation to select high-quality training data; algorithms based on data distribution information, which optimize sample selection strategies based on the characteristics of data distribution; algorithms for optimizing model predictions, including methods for optimizing the acquisition and utilization of deep model prediction information, improving the predictive model structure through the use of generative adversarial networks and reinforcement learning, as well as enhancing model prediction performance based on the Transformer architecture to ensure the reliability of model predictions. In addition, this paper also conducts experimental comparisons on important academic work under various types of active learning image classification algorithms, and analyzes the performance and adaptability of each algorithm on datasets of different scales. Furthermore, this paper discusses the challenges faced by active learning image classification technology and points out future research directions.

Federated learning is a distributed machine learning paradigm that facilitates data sharing and collaborative computing among multiple participants. Currently, research on federated learning primarily focuses on performance improvement and privacy protection. With the emergence of trustworthy artificial intelligence, the research on trustworthy federated learning methods has gained more attention, and ensuring fairness in federated learning is one of the main challenges. Improving the fairness of federated learning can motivate the enthusiasm of clients and ensure the sustainability of federated learning training. However, due to the heterogeneity of data and devices in federated learning, traditional federated learning methods may lead to significant performance differences between clients, which may hinder fairness among all participants and significantly impact the motivation of users to participate in federated learning. Based on this, this paper provides a comprehensive review of the research methods of fairness in federated learning. Firstly, we categorize the main research directions of fairness in federated learning, elaborates the definition and compares the evaluation criteria of fairness in each direction. Next, we discuss the challenges and main solutions for improving fairness in federated learning in each direction. Then, we summarize the commonly used datasets, experimental scenarios, and fairness evaluation metrics in the study of fairness. Finally, we prospectively explore the future research directions and development trends of fairness in federated learning.

In recent years, significant progress has been made in multi-exposure image fusion in dynamic scenes. In particular, the deep learning based methods have shown great visual performance in dynamic multi-exposure image fusion, which have become the mainstream methods in high dynamic range (HDR) imaging. However, the current deep learning based methods are mostly implemented in a supervised manner, which heavily rely on the ground-truth images. That makes it difficult for them to work in real scenes. In this paper, we propose a self-supervised multi-exposure image fusion network for dynamic scenes. The main contributions of this paper are as follows: we design a self-supervised fusion network to explore the latent relationship between HDR and low dynamic range (LDR) images; we propose an attention mechanism based global deghosting module, to reduce the ghosting artifacts caused by moving objects; we propose a merging reconstruction module with residual and dense connections, to improve the reconstruction details; we design a motion mask guided self-supervised loss function to train the proposed network efficiently. Experimental results demonstrate the effectiveness of the proposed method. Compared with the state-of-the-art methods, our method achieves higher objective and subjective quality on reconstructed HDR images, with faster running speed.

In this paper, a broadband low-profile dual-polarized cross-dipole antenna loaded with a metasurface (MS) is proposed. The antenna is consisted of three parts: a pair of crossed dipoles, four parasitic patches, and a metasurface structure. The crossed dipoles are used to achieve dual-polarization characteristics of the antenna. By loading parasitic patches above the dipoles and slotting the dipole arms, the impedance bandwidth of the antenna is extended, and the low profile of the antenna is achieved by replacing the metal reflector below the dipole with a metasurface. In order to improve the isolation between the input ports, four metal shorting columns are introduced. The simulation and measured results show that the impedance bandwidth of dB is 42.5% (2.26~3.48 GHz), and the port isolation and cross-polarization within the bandwidth range are greater than 21 dB and below -31 dB, respectively. The size is only \mathrm{0.5} {\lambda }_{\mathrm{0}} \times \mathrm{0.5} {\lambda }_{\mathrm{0}} \times \mathrm{0.074} {\lambda }_{\mathrm{0}} ( \lambda ₀ is the wavelength of free space corresponding to the operating frequency of 2.9 GHz).

Camera and lidar are the key sources of information in autonomous vehicles (AVs) . However, in the current 3D object detection tasks, most of the pure point cloud network detection capabilities are better than those of image and laser point cloud fusion networks. Existing studies summarize the reasons for this as the misalignment of view between image and radar information and the difficulty of matching heterogeneous features. Single-stage fusion algorithm is difficult to fully fuse the features of both. For this reason, a nova 3D object detection based on multilayer multimodal fusion (3DMMF) is presented. First, in the early-fusion phase, point clouds are encoded locally by Frustum-RGB-PointPainting (FRP) formed by the 2D detection frame. Then, the encoded point cloud input is combined with the self-attention mechanism context-aware channel to expand the PointPillars detection network. In the later-fusion phase, 2D and 3D candidate boxes are coded as two sets of sparse tensors before they are not greatly suppressed, and the final 3D target detection result is obtained by using the camera lidar object candidates fusion (CLOCs) network. Experiments on KITTI datasets show that this fusion detection method has a significant performance improvement over the baseline of pure point cloud networks, with an average mAP improvement of 6.24%.

The flexible sensor based on photo plethysmo graphy (PPG) can detect heart rate (HR) and blood pressure (BP), but the calibration of their detection results is rarely reported. Therefore, this paper proposes a reflective PPG heart rate detection and blood pressure calibration system based on simulated blood circulation.The peristaltic pump is used to generate pulsating flow,and the frequency and pressure of simulated blood delivery are controlled by adjusting its rotational speed,thus causing the change of the volume of simulated blood in the elastic latex tube and changing the signal period and intensity of reflected light, which is closer to match the actual scenario of human pulse measurement process.The mean value of heart rate detection error of the system is 0.277 78, and the 95% consistency limit is (-2.595 62, 3.151 17). The goodness of fit of measured systolic blood pressure (SBP) and diastolic blood pressure (DBP) are 0.971 85 and 0.981 11, respectively. The mean value of mean deviation (MD) ± standard deviation (SD) of SBP and DBP detected by the calibrated flexible PPG sensor on four volunteers is (1.21±2.16) mmHg and (0.76±2.02) mmHg, respectively, which are in line with and far less than the standard index of 5±8 mmHg for measuring the accuracy of blood pressure monitors set by the association for the advancement of medical instrumentation (AAMI). The results show that this system can calibrate the flexible PPG sensor accurately and efficiently, which provides the calibration basis for the accurate blood pressure detection of portable wearable devices.

The existing object detection methods are insufficient for low-light images due to their intrinsic property such as low contrast, detail loss and high noise. To solve this problem, a low-light object detection method that combines spatial-aware attention mechanism with multi-scale feature fusion (SAM-MSFF) is proposed. Firstly, multi-scale features are fused by multi-scale interactive memory pyramid to enhance effective information under low-illumination condition, and features of memory vector storage samples are set to capture potential correlation between samples. Then, a spatial-aware attention mechanism is introduced to obtain long-distance context information and local information of features in spatial domain, thereby enhancing the object features in low-light images and suppressing the interference of background information and noise. Finally, multiple receptive field enhancement module is used to expand receptive field of the features, and the features with different receptive fields are grouped and re-weighted, so that detection network can adaptively adjust the size of receptive field according to input multi-scale information. Experimental results on the ExDark dataset show that mAP (mean Average Precision) of the proposed method reaches 77.04%, which is 2.6%~14.34% higher than existing mainstream object detection methods.

Images captured in low-illumination environments often have many quality problems, such as weak brightness, low contrast, much noise, and detail loss. These problems will lead to inaccurate localization and object classification errors when using the existing object detection models to detect low-light images, resulting in low detection accuracy. Aiming at the above phenomena, this paper proposes a low-illumination object detection method called Night-YOLOX. First, the low-level feature gathering module (LFGM) is designed to be incorporated into the backbone. Capturing more effective low-level features in low-illumination scenes is beneficial to locating objects. The LFGM aggregates more discriminative low-level features from the shallow feature maps and feeds them into the high-level feature maps and the deep convolution stages, so as to compensate for the loss of low-level edge, contour, and texture features during feature extraction in low-light images. Then, the attention guidance block (AGB) is designed to be embedded in the neck of the detection model. The AGB reduces the influence of noise interference in low-light images, guides the detection model to infer the complete object regions and extract more useful object feature information, so as to improve the accuracy of object classification. Finally, experiments are conducted on the real low-light image dataset ExDark. The experimental results show that compared with other mainstream object detection methods, the proposed Night-YOLOX has better detection performance in low-illumination scenarios.

Over-the-air computation (AirComp) is an effective method to improve the efficiency of distributed data aggregation, which can complete some task calculations while transmitting in the air. Most existing researches focus on the single unmanned aerial vehicle (UAV) scheme, without considering the quality of data aggregation and the stability of the system, making it unsuitable for practical AirComp environments. Therefore, this paper proposes an AirComp network based on multiple UAVs collaboration, which aims to achieve the efficient data aggregation for multiple ground mobile sensors (GMSs). In order to refine data acquisition and fully reflect system status, a multi-constraint non-convex optimization problem is constructed to jointly optimize UAV-GMS association, the three dimensional (3D) deployment of UAVs, UAV denoising factors, and transmission power allocation, aiming for maximizing the system's minimum achievable rate. Giving the nonlinear characteristics of multiple constraints optimization problems, a deep deterministic policy gradient-based optimization algorithm for multiple UAVs cooperation in AirComp network (AirDDPG-UAV) is proposed to assist UAVs rapidly responding to aggregation missions in complex environments. A deterministic policy in deep reinforcement is adopted to optimize the states, behaviors, and rewards of the AirComp network, aiming to maximize the minimal achievable rate. The numerical results show that the AirDDPG-UAV algorithm can significantly improve the system's minimum achievable rate by more than 15% compared to the benchmark methods, while ensuring suitable system energy consumption and computational complexity. The AirDDPG-UAV algorithm also obtains satisfactory results in optimizing the mean MSE, which illustrates our method has excellent performance in scaling signals and thus is helpful for fast data aggregation. The experiments indicate the proposed scheme is appropriate for the distributed data aggregation with low cost and can obviously improve the efficiency and stability of data aggregation.

The Internet, as a critical component of a nation's information infrastructure, has played a significant role in various domains. However, as its scale continues to expand and its applications deepen, we also face the potential catastrophic consequences of inconsistent network behaviors. To ensure the normal operation of the Internet and the consistency of network behaviors, there is an urgent need for deployable network verification technologies that align network operations with the intentions of network operators. Extensive research has been conducted on network verification technologies, assisting users in automating the detection of network errors and analyzing their root causes. However, to meet the increasing demands of the expanding Internet, scalability has become a crucial challenge in deploying network verification technologies. Specifically, how to quickly identify and diagnose errors in network policies, while satisfying time and space complexity constraints, has become a research hotspot in effectively applying network verification technologies in practice. To address this problem, this paper delves into and summarizes cutting-edge research on the temporal and spatial scalability of network verification. It begins by introducing the background knowledge related to network verification and then describes the current issues and challenges faced in network verification. Focusing on the core issue of scalability, the paper thoroughly analyzes existing work in achieving scalable verification from both the data plane and control plane perspectives. It provides a systematic analysis of the characteristics of these approaches, showcasing the distinctions and connections among related studies. According to the existing researches, we find that: (1) The scalability of data plane verification is primarily constrained by header space and forwarding matching rules, while the scalability of control plane verification is mainly limited by the complexity of multiple protocols and policies. (2) Although both data plane and control plane research employ similar scalable verification techniques, they address different but interconnected targets. For example, incremental computation in the data plane primarily focuses on updating packet equivalence classes, while incremental computation in the control plane primarily deals with network models affected by configuration changes. When applying network slicing techniques, both data plane and control plane independently validate the network by dividing it into multiple segments. (3) Compared to spatial scalability, current research places greater emphasis on temporal scalability, where reducing verification time overhead appears to be the primary pursuit of verification tools. (4) Previous research predominantly adopted a centralized verification approach, which involved collecting control plane or data plane information and then performing centralized analysis and verification. However, there has been a recent trend towards distributed verification, such as Coral and Tulkun in control plane verification. Lastly, based on the current research landscape, the paper concludes by summarizing and forecasting the research trends in scalable network verification technologies, offering valuable insights for researchers in this field. In conclusion, this paper presents a comprehensive review and outlook on the topic of scalability in network verification. It emphasizes the importance of aligning network behaviors with the intentions of network operators to ensure the reliable and consistent operation of the Internet. By addressing the challenges of scalability, researchers can advance the development of network verification technologies that can effectively verify large-scale networks within the constraints of time and space complexity. Ultimately, this contributes to enhancing the reliability and security of the Internet as a critical information infrastructure.

As an emerging distributed machine learning architecture, federated learning (FL) allows multiple users to train local models and achieve global aggregation of models with data privacy protection, thus providing reliable Internet of Vehicle (IoV) services. However, in the training process of FL, many training terminals may switch among domains due to the high mobility, resulting in low accuracy of the global model. Besides, malicious terminals may frequently upload invalid or incorrect model data which leads to low service reliability. Therefore, we build the dual-layer FL based edge collaborative computing mechanism for high dynamic IoV businesses. Firstly, we comprehensively consider the mobility, computing ability and reliability to construct the service capability model for the terminal, and then propose the edge collaborative computing domain (ECCD) construction algorithm based on deep reinforcement learning. By clustering the vehicle terminals covered by multiple edge nodes, the switching probability of the terminal local model will be reduced, and the sustainability of the FL model training can be guaranteed. Furthermore, we design a dual-layer FL framework including the inter-ECCD aggregation layer and cross-ECCD aggregation layer, respectively. It adopts the semi-asynchronous aggregation mechanism for local models based on the adaptive aggregation factor in the inter-ECCD aggregation layer, and the asynchronous aggregation mechanism for domain’s regional model based on data volume in the cross-ECCD aggregation layer, which jointly improve the aggregation efficiency of the FL system. In particular, considering that the high speed terminals may cause the cross-domain problem, we introduce the partial conditional update mechanism for the local model to avoid the situation that the high-quality models are covered by the low-quality models, which further improves the accuracy of the global model and the utilization of FL system resources. The simulation results verify that the proposed framework outperforms the local computing and asynchronous/synchronous FL algorithms in terms of model accuracy and service reliability.

With the implementation of major national strategies in industries such as intelligent manufacturing and transportation, determinism has become a new focus of information networks, especially industry-specific networks. Existing deterministic network technologies provide deterministic guarantees based on network transmission elements (e.g., bandwidth or time slots). However, relying solely on network transmission elements does not support the diverse needs of emerging industry applications. For example, in computing network integration scenarios, intelligent computing tasks require the determinism of transmission and computing elements to achieve high-performance communication. In green communication scenarios, the determinism of node energy elements needs to be considered to maintain network operation stability. In response to the above requirements, this paper studies generalized deterministic identification networks with respect to multiple elements such as transmission, computing, storage, and energy based on a previously proposed network identification technology. First, a generalized deterministic identification network architecture is proposed that includes a differentiated service layer, a heterogeneous network layer, and an intelligent adaptation layer. The differentiated service and heterogeneous network layers uniformly identify the deterministic applications and networks. The intelligent adaptation layer schedules the network resources in units of flow. Existing deterministic resource scheduling methods, even if they only consider the basic deterministic elements in a single network, still face problems such as long computational time, high complexity, and low flexibility. To support a more complex collaborative adaptation of multiple deterministic elements, the end-to-end deterministic resource scheduling (E2eDet) algorithm, which is based on deep reinforcement learning, is designed. To meet the various deterministic requirements of different applications, E2eDet uniformly and collaboratively allocates multiple deterministic network resources for mixed data streams from end to end. Experimental results show that E2eDet increases the amount of data flow scheduling by 28.4% and 6.38× when compared with the DeepCQF and Random algorithms, respectively. Moreover, E2eDet can better balance the computational time and scheduling ability.

In order to facilitate scientific researchers to understand the project application, acceptance, and funding status of the “Semiconductor Science and Information Devices” discipline direction of the National Natural Science Foundation of China, this article conducts a statistical analysis of the status of the projects in 2023. Firstly, the important reform measures of the National Natural Science Foundation of China in 2023 is briefly introduced. Subsequently, the application and funding status of F04 including general projects, youth science fund projects, regional science fund projects, key projects, outstanding youth science fund projects, and national outstanding youth science fund projects are summarized and analyzed. The distribution of supporting organizations of the general program, youth science fund projects, regional science fund projects are analyzed, as well as the four types of scientific problem attributes of the applied projects. Finally, the priority development direction in the field of “semiconductor science and information devices” is prospected.

The utilization of the synergistic fusion of low resolution hyperspectral image (LR-HSI) and high resolution multispectral image (HR-MSI) for the purpose of achieving enhanced hyperspectral spatial resolution has emerged as a prominent and actively pursued research area within the domain of hyperspectral image processing. At the present time, deep learning has become an efficient tool for HSI-MSI fusion. Despite the potential of deep learning, there are still some challenging, such as how to effectively mine the complementary information of HSI and MSI, how to inject the spatial structure and detail of MSI into HSI, and how to maintain the spectral fidelity of HSI. This study proposes a multilevel wavelet-deep aggregation network (MW-DAN). It has dual branches, which combine undecimated wavelet transform (UDWT) with deep residual network to promote the image reconstruction. Particularly, the UDWT directional subband decomposition of MSIs is performed by introducing jumper aggregation connections in the deep residual network to design an information aggregation type structure, and injected into the middle hidden layer of the network layer by layer to enhance the detail injection and spectral fidelity of the directional subband structure. The entire network are trained from LR-HSI, HR-MSI and HR-HSI in an end-to-end fashion. It could learn the spatial-spectral fusion nonlinear mapping with superior performance. Experimental results on simulation and real datasets show that the proposed method is superior to the state-of-the-art fusion methods in terms of objective evaluation index, spectral fidelity and visual performance.

Aiming at the problems of unbalanced edge device resources, communication delay and low model quality in the field of environmental monitoring, this paper proposes an adaptive federated learning algorithm for environmental monitoring based on edge computing. This algorithm aims to use edge devices for data processing, and according to each the resource limitation of the device adjusts the aggregation frequency of the global model to better adapt to different monitoring environments. By considering the resource differences between edge devices, the algorithm adopts a strategy of dynamically optimizing the iteration frequency to improve the training effect of the model. Compared with the traditional fixed iteration frequency, the adjustment strategy of this algorithm is more flexible and can better adapt to different data distribution and participant characteristics. Through a large number of experimental evaluations, and using the same algorithm convolutional neural networks-federated learning (CNN-FL), federated averaging (FedAvg) and hierarchical federated edge learning (HFEL), the algorithm proposed in this paper has significant advantages in algorithm performance and economic cost. This algorithm provides an efficient, safe and reliable method for environmental monitoring. Expanded approach to data analysis and modeling to help drive improvements in environmental monitoring capabilities.

In order to locate radio silent target on the Earth surface, a novel positioning method for moving target is proposed based on passive radar system. Unlike most existing positioning methods, this method takes into account the quadratic constraints that the target position vector and velocity vector need to satisfy, and can achieve decoupled estimation of the target position and velocity. Firstly, the nonlinear observation equations based on the passive radar system are transformed into pseudo-linear observation equations. Then, the asymptotic statistical properties of the errors in pseudo-linear equations are derived by applying the first-order error analysis. Subsequently, an optimization criterion is constructed for the joint estimation of the target position and velocity under the two quadratic equality constraints. For the purpose of obtaining the global optimal solution of the target position and velocity, a decoupled optimization algorithm is developed based on the Lagrange multiplier approach. This optimization algorithm requires only the iteration of the target position parameters, and the target velocity parameters can be obtained in a closed form, thus reducing the influence of initial values and the risk of local convergence. Furthermore, the Cramér-Rao bound (CRB) for moving target localization based on passive radar system is deduced under the two quadratic equality constraints, and the performance gain resulting from the equality constraints is quantified. The new estimator is also proved to be asymptotically statistically efficient by using the first-order error analysis as well as the Lagrange multiplier approach. Finally, simulation results verify the advantages of the proposed positioning method.

Person re-identification (re-ID) aims to identify a person's images across different cameras. However, the domain bias between different datasets makes it a challenge for re-ID models trained on one dataset to be adapted to another. A variety of unsupervised domain adaptation methods tend to transfer learned knowledge from one domain to another by optimizing with pseudo-labels. However, these methods introduce a large number of noisy labels through one-shot clustering, which hinders the retraining process and limits generalization. To mitigate the impact of noisy pseudo-labels, this paper proposes an unsupervised person re-identification method based on an ensemble of heterogeneous convolutional neural networks. The framework does not apply any manual labeling information, automatically infers the relationship between pedestrian images in the target domain, and a cooperative trusted instance selection mechanism is established to select pseudo-labels with high credibility. By constructing a dual-branch heterogeneous network, a variety of different pedestrian features are learned, and memory structures are designed to store the life-long features during the training stage, which could reduce the fluctuation of noise labels, and improve the robustness of the model. Comprehensive experimental results have demonstrated that our proposed method can achieve excellent performances on benchmark datasets. And mAP is increased to 85.4% and 74.8% on Market1501 and DukeMTMC-reID, respectively.

In the field of airborne communications, there is an urgent need for a low-profile miniaturized antenna with vertically polarized omnidirectional radiation patterns, especially in VHF and UHF bands. To satisfy this requirement, a cylindrical cavity antenna with a low profile and electrically small size is proposed in this paper. By introducing two short-circuited columns and a conductive column into the cylindrical cavity to excite the TM₀₁ and TM₀₂ modes, respectively, good vertical polarization omnidirectional radiation patterns can be obtained. A tapered sleeve is introduced to improve the input impedance and effectively reduce the resonant frequencies of the TM₀₁ and TM₀₂ modes so that the resonant frequencies of the two modes can work in the target frequency band. Besides, a broadband matching network with four states is designed to further improve the impedance matching of the proposed antenna. The size of the proposed cylindrical cavity antenna is \mathrm{\pi } \times ( \mathrm{0.11} {\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}} {)}^{\mathrm{2}} \times \mathrm{0.045} {\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}} ( {\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}}is the free space wavelength at the lowest frequency). There is good agreement between the simulated and measured results. The measured results show that the proposed cylindrical cavity antenna achieves 47.3% (105~170 MHz) and 52.3% (207~355 MHz) impedance bandwidth in the low-frequency band and high-frequency band ( ), respectively, and a stable vertical polarization omnidirectional radiation pattern is demonstrated in the wide frequency band.

Recently, the outstanding text generation language models represented by ChatGPT, which can adapt to complex scenes and meet various application demands of human beings, has become the focuses of both the academic and industrial circles. However, the advantage of large language models (LLM) such as ChatGPT that are highly faithful to user intent implies some factual errors, and it is also necessary to rely on prompt content to control the detailed generation quality and domain adaptability, so it is still of great significance to study text generation with intrinsic quality constraints as the core. Based on the comparative study of key content generation models and technologies in recent years, this paper defined the basic form of text generation with intrinsic quality constraints, and six quality features based on “credibility, expressiveness and elegance”. In view of these 6 quality features, we provided analysis and comparison of generator model design and related algorithms. Besides, various automatic and human evaluation methods for different intrinsic quality features are summarized. Finally, this paper looks forward to the future research directions of intrinsic quality constraint technology.

The electric power infrastructure of China has developed into a highly informationized, automated, and intelligent cyber physical integration system. The interaction of cyber and physical not only significantly improves the efficiency and performance of power supply, but also introduces new network security threat. Cross-domain attacks that occurring in the cyber domain and acting on the physical domain can cause the systematic breakdown of power infrastructure and then lead to large-scale power outages. However, the current isolated cyber side or physical side defense system is difficult to effectively deal with these cross-domain attack threats.This paper introduces the current situation of information and physical cross domain attack threats faced by the power system, elaborates on the shortcomings of traditional defense methods in facing cross domain attacks, proposes a cross domain attack defense architecture based on information and physical collaboration, and designs defense methods from the perspectives of perception, identification, and blocking on the attack time scale. Through example design, it is proven that the proposed information and physical collaboration defense architecture can ensure the safe and stable operation of the power system.

To solve the problem of multi-channel cross-talk in full-space metasurface, a strategy of tri-band transmission-reflection-transmission full-space wavefront control based on dual-geometric-phase metasurface is proposed. A frequency selective surface was cascaded with multilayer metasurface to obtain independent modulation of three frequency bands. The full-space independent amplitude and phase control combining transmission and reflection was realized based on dual geometric phase theory. For verification, a full-space multifunctional metadevice working at C, X and Ku bands was designed, fabricated and measured. Both numerical and experimental results show that transmissive and reflective dual-vortex beams along y-axis and x-axis are generated at 7 GHz and 10.2 GHz, respectively, meanwhile dual foci along x-axis is generated at 15.7 GHz under left-handed circularly-polarized wave excitation. The proposed strategy of tri-band transmission-reflection-transmission full-space wavefront control provides new approaches for new multifunctional devices and integrated electromagnetic wave control, and expands the application prospect of metasurface in large-capacity communication.

This paper analyzes the existed processing scheme, and proposes a multi-scale semantic encoder-decoder networks (MSEDNet) by comprehensively using multiple technologies for the problems in remote sensing image semantic segmentation both multi-level information extraction and multi-scale feature diagram dependence characteristic. The MSEDNet consists of two parts: encoding part and decoding part. In the encoding part, the enhanced MobileNetV3 with residuals coordinate spatial attention (RCSA) is firstly proposed to extract semantic information, and then a multi-layer enhanced semantic context module (ESCM) is designed to improve representation ability of the multi-scale structure feature map. In the decoding part, a strengthen spatial detail information module (SSDIM) based on Multi-core Convolution and Focus Parallel is proposed to enhance the details and structural information of shallow features. Then triplet iterative multi-scale feature fusion (TIMSFF) strategy is designed to strengthen the multi-scale context fusion both deep global semantic information and shallow local detail features, for improving the segmentation accuracy. The proposed model has been experimentally verified on the ISPRS Vaihingen and Potsdam dataset. The overall segmentation accuracy (OA) reached 95.699% and 95.534% respectively, the mean F ₁-score (mF ₁) increased by 2.661% and 2.929% respectively, and the mean intersection over union (mIoU) increased by 3.973%and 4.012%, respectively. The number of param dropped to 6.77 M.

The 3D object detection of camera and lidar multimodal fusion can comprehensively utilize the advantages of the two sensors to improve the accuracy and robustness of detection. However, due to the complexity of the environment and the inherent variability among multimodal data, 3D object detection still faces many challenges. In this paper, we propose a multimodal 3D object detection algorithm with a double-fusion framework. We design a voxel-level and grid-level double-fusion framework, effectively alleviating the semantic differences between modal data. We propose the ABFF (Adaptive Bird-eye-view Features Fusion) module to enhance the algorithm's ability to perceive small object features. Through voxel-level global fusion information to guide grid-level local fusion, we propose a Transformer-based multimodal grid feature encoder to extract richer context information in 3D detection scenes and improve the efficiency of the algorithm. The experimental results on the KITTI standard dataset show that the average detection accuracy of our proposed 3D object detection algorithm reaches 78.79%, which has better 3D object detection performance.

Alzheimer's disease (AD) is a degenerative disease, as the disease worsens, the patient's language ability gradually decreases. Some researchers have already used acoustic features such as Mel spectrogram and Mel frequency cepstral coefficient (MFCC) to classify AD patients and healthy individuals, but there is a lack of further exploration on using neural networks to extract features from raw waveforms for AD detection. In this paper, we propose an end-to-end AD detection method based on raw waveforms. The method uses one-dimensional convolution to extract time-dimensional features from the original waveform and uses a residual block containing an inflated convolution to extract more complex features. To further improve performance, the squeeze-and-excitation block is introduced into the residual block. On the national conference on man-machine speech communication (NCMMSC) 2021 AD dataset, the model proposed in this paper achieves 86.55% and 81.35% accuracy on the long audio test set and short audio test set, respectively, which is 6.75% and 7.35% better than the baseline system, respectively. On the INTERSPEECH2020 ADReSS dataset, the accuracy of the model is 66.67%, an improvement of 4.17% over the baseline system.