This PDF file contains the front matter associated with SPIE Proceedings Volume 13222, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Image hiding is the technique of embedding secret information into a carrier in a covert manner to ensure secure transmission of information. Traditional image hiding has insufficient resistance to the latest steganalysis tool based on deep learning and has low robustness. In addition, current image hiding models have relatively lower hiding capacity and mostly only hide gray images. This paper proposed steganography based on generative pixel shuffle residual network (GPRS-Net), that is capable of hiding single or multi color secret images in a carrier image. The secret image is decomposed and encoded into different feature maps using a preprocessing network, which allows the GPRS-Net to better fuse the secret information into the carrier image, thus enabling the output of the stego image. Subsequently, a deep recovery network based on residual module and pixel shuffling module is constructed, with the aim of potentially recovering the secret image with less loss from the attacked stego image, thus improving robustness and security. Experiments show that the proposed GPRS-Net has excellent performance in terms of hidden capacity and security.

In this paper, we propose an efficient joint precoding design method to maximize the weighted sum-rate in Intelligent Reflecting Surface (IRS)-assisted cell-free networks by jointly optimizing the active beamforming of base stations and the passive beamforming of IRS. To address the high-dimensional non-convex optimization problem, we employ a fractional programming approach to decouple the non-convex problem into subproblems for alternating optimization between active and passive beamforming. The active beamforming subproblem is addressed using the Consensus Alternating Direction Method of Multipliers (CADMM) algorithm, while the passive beamforming subproblem is tackled using the Accelerated Projection Gradient (APG) method. Simulation results demonstrate that our proposed approach achieves significant improvements in weighted sum-rate under various performance metrics compared to Primal-Dual Subgradient (PDS).

In the high-speed mobile environment supported by the fifth-generation mobile communication technology, higher vehicle speeds, more frequent switching and wider bandwidth make the design of high-speed mobile communication systems more challenging. High-quality wireless communication technology is needed to support future high-speed mobile scenarios to achieve low-latency and highly reliable communication, among which anti-Doppler frequency shift technology is the key. Based on this, this paper proposes a Rayleigh channel estimation algorithm based on image super-resolution (SR) network. This method models the channel response matrix obtained by the LS method as a two-dimensional image, uses the SR network and image restoration (Image restoration, IR) technology for learning optimization, and obtains the channel response at the unknown symbol. The simulation results show that the proposed algorithm is significantly better than the traditional Linear Minimum Mean Square Error (LMMSE) method. Compared with other SR-based channel estimation algorithms under low Signal to Noise Ratio (SNR), it has a 3dB gain and a 5dB gain under high SNR.

Similar duplicate record recognition can help us effectively remove duplicate and redundant data, improve data quality and consistency. To this end, a domestic database migration similar duplicate record recognition method based on an improved LSSVM algorithm is proposed. Extract domestic database migration data based on crawler structure; Reduce noise and encode the extracted migration data separately to improve data quality; Based on the time cycle, using mutual information as a correlation measure, extract and fuse duplicate features of transfer data. Construct a similar duplicate record recognition model using the LSSVM algorithm and improve the LSSVM algorithm using the quantum particle swarm optimization algorithm to achieve similar duplicate record recognition in domestic database migration. The experimental results show that the method has high recognition ROC and F1 values, which improves recognition accuracy and efficiency.

In advancing Neural Style Transfer (NST) for video applications, this study introduces a novel Transformer-based framework that enhances temporal consistency and computational speed. The approach utilizes a Vision Transformer (ViT) architecture, adapted to exploit the temporal dynamics inherent in video sequences. To achieve seamless style consistency across frames while ensuring real-time performance, we incorporate a hybrid Transformer-Residual learning module coupled with a bi-directional frame recurrence strategy. Our DVT: Dual-frame Video Transfer also integrates advanced model acceleration techniques, including network pruning and knowledge distillation, to foster an efficient processing environment capable of real-time stylization. Through rigorous experimental validation on diverse video datasets, our model outperforms current NST techniques, delivering superior stylization quality, improved temporal stability, and heightened computational efficiency. The findings highlight the pivotal components of our architecture.

The high integration of FL to achieve data privacy preservation in modern mobile applications is bringing users’ data to become the most valuable assets in the challenge of realizing robust data privacy policies inside network infrastructures. This means, in other words, that today, securing sensitive user data is becoming more essential than before. In this article, by using FL based privacy preserving ecosystem learn from the situation of data centralized context, we adopt the differential data sharing to realize high performance inside FL blockchain by increasing also its preservation performance. Our result performances by comparing privacy and non-privacy preserving chaincode FL Blockchain shows high performance on keeping protected the data shared environment. Hyperledger Fabric algorithm design simulation scenario reduces considerably all addition noise to keep the entire system work properly. This contribution shows how mobile communication network systems in data privacy and preservation context can be incentivized to allow any users through interchange connection to be rewarded for their work performances, and how differential data sharing can be useful to deal with it.

This paper addresses the challenges faced by towed array sonar systems, including near field interference from platform noise and beamforming distortion from array irregularities. A novel approach is proposed, combining spatial matrix filtering and plane wave azimuth estimation technology. By leveraging the characteristics of near field platform noise and far field target signals, an array data reconstruction filter is designed to mitigate interference before target azimuth estimation. This method aims to enhance target detection performance by suppressing platform noise and improving array data reconstruction.

With the rapid advancement of industrialization and the Internet of Things, "Industry 4.0" will propel industrialization and informatization to a new stage of development simultaneously. The MQTT protocol, as a lightweight transmission protocol widely utilized in the industrial Internet of Things transmission environment, faces stringent security requirements for industrial data transmission. Traditional use of the MQTT protocol may result in identity authentication deficiencies and plaintext message content transmission, among other security issues. To address these concerns, this paper proposes an MQTT-DH encryption scheme based on symmetric encryption. By incorporating the Diffie-Hellman algorithm into the process of establishing connections between MQTT clients and servers, communication parties can generate a shared secret key to verify each other's identities. Furthermore, to enhance security post-MQTT connection establishment, session counters and shared secret keys are introduced as initialization vectors for AES-CBC encryption algorithms. This enhancement significantly improves data security levels. Experimental results demonstrate that the proposed encryption scheme maintains excellent performance and high security across numerous data encryption tests while providing efficient and secure services for industrial data transmission.

Due to the different action modes of communication network attacks, it is difficult to guarantee the effect of defense technology in information loss control. Therefore, a quantum secure communication network attack defense method based on AES algorithm and time window is proposed. AES algorithm is introduced to group the original quantum secure communication data plaintext to obtain the corresponding data blocks. After multi-level conversion processing, the intermediate results generated by each conversion are taken as the state parameters, and data encryption is completed through byte substitution operation, row transformation operation, column mixing operation and round key transformation operation. When defending against quantum secure communication network attacks, the network state decision time window is established. From the perspective of the traffic status in the window, the attack status and location are analyzed and judged, and the dynamic adjustment of the traffic limit strategy is adopted for attack defense. In the test results, the communication packet loss rate is the lowest, basically stable within 3.0%, which has obvious advantages compared with the control group.

As optical fiber communication technology develops towards high speed, large capacity and long distance, the impact of optical fiber nonlinearity and other damages in optical fiber communication systems is becoming more and more serious. In order to ensure the reliability and stability of the system, Optical Performance Monitoring (OPM) technology is of vital importance. It's important. At present, a single Optical Signal-to-Noise Ratio (OSNR) monitoring technology is difficult to meet the requirements of system operation and maintenance. Joint monitoring of OSNR, nonlinearity and modulation format is an important development direction of OPM. This paper studies multi-parameter joint monitoring technology of OSNR, nonlinear noise power and modulation format based on different deep learning models. Asynchronous Complex Histogram (ACH) is designed as the input feature data of the neural network, and the MT-CNN joint monitoring scheme based on AAH is designed. The monitoring results show that the joint monitoring scheme reduces the total number of learnable parameters and saves computing resources and memory overhead. For QPSK, 16QAM, and 32QAM signals, high-precision modulation format identification can be achieved. The root mean square error of nonlinear noise power and OSNR monitoring values reaches 0.21 dB and 0.22 dB respectively. It can also realize changing transmission distance and emission The robustness of power and laser linewidth is superior in performance and computational efficiency; a low-complexity optical fiber nonlinear damage compensation scheme based on a simple recurrent neural network is studied. This compensation scheme significantly reduces the number of triple products required for the equalization process, which can reflect the nonlinear damage characteristics of the optical fiber, by taking advantage of the time memory of the recurrent neural network.

Convolutional Neural Networks (CNN) are widely utilized in the field of computer vision, garnering particular attention in facial emotion recognition. Literature analysis indicates that attention mechanisms significantly enhance model precision and accuracy. However, existing facial emotion recognition models have not fully considered the interplay between convolutional neural networks and machine vision, nor the correlation between facial emotion recognition and the target user groups. This paper proposes a novel model, MD-CNN, which integrates convolutional neural networks with an attention mechanism and selects appropriate datasets based on the characteristics of the target groups. The model design includes an attention layer and experiments are conducted using the Affect Net dataset. Comparisons with other CNN emotion recognition models demonstrate that the proposed model significantly improves accuracy.

The Generalized Sidelobe Canceller (GSC) is a classical beamforming technique widely used in microphone array speech enhancement. To enhance its interference suppression capability in complex acoustic environments, this paper integrates the Spatial Sound Presence Probability (SSPP) based on the Relative Transfer Function (RTF). By estimating the spatial presence probability of the target sound source using RTF characteristics, the method distinguishes between target and interference sources in the frequency domain. This probabilistic reconstruction of the target sound source ultimately improves speech quality. Simulation results demonstrate that the improved GSC exhibits superior speech quality and interference suppression under various Signal-to-Interference Ratio (SIR) conditions.

As the application of Internet of Things (IoT) technology in agriculture deepens, the lack of security in agricultural IoT systems has become one of the main bottlenecks restricting the development of agricultural IoT. To address the data and system security issues of agricultural IoT, this paper proposes an agricultural IoT PDRR scheme based on an improved cyber kill chain. By understanding the lifecycle of network attacks through the improved four-step cyber kill chain model, and utilizing the synergistic effects of honeypot and firewall technologies, a defense system capable of adapting to attacker behaviors and effectively disrupting the cyber kill chain is constructed. The PDRR is achieved by combining proactive and passive defense methods. Experimental results show that the agricultural IoT PDRR scheme based on the improved cyber kill chain proposed in this paper can effectively interfere with and disrupt the cyber kill chain. Moreover, compared to traditional defense methods, this scheme demonstrates significant advantages in recognizing unknown attack patterns and acquiring attack intelligence.

With the continuous improvement of economic level, the number of cars in cities shows a trend of continuous growth. Internet of Vehicles, as the foundation layer of smart cities, can greatly improve transportation efficiency, reduce energy consumption, and reduce traffic accidents. Currently in order to cope with the security and privacy issues of the Internet of Vehicles, researchers have proposed a variety of authentication schemes and agreement protocols, however, existing authentication protocols and key agreement protocols are mostly split apart, so it is necessary to devise a secured and reliable authentication key agreement protocol scheme to protect the communication channel in the Internet of Vehicles. To address the problem that vehicles in different regions need to communicate, we propose a cross-domain certificateless based authentication key agreement protocol. The certificate-less regime is used to achieve lightweight and efficient authentication key agreement, solve the certificate management and key escrow problems, and leverage the blockchain to introduce a decentralized KGC architecture for this protocol to enable cross-domain authentication. We conclude that we gain significant efficiency advantages with respect computational expense and communication expenditure while ensuring the necessary security properties.

Early fire Due to the small target and the occurrence of early scene complexity, the existing fire detection methods make it very easy to miss the detection of false detection phenomena. To address the above problems, this study designs CAC-Net, a fire detection method based on residual feature extraction and a self-attention mechanism. The method utilizes an efficient two-step residual feature extraction module to finely tune the receptive field and enhance the extraction of complex features as well as small object features. Effective Pyramid Aggregation Module AF accurately aggregates task-related information. The CGD module uses group convolution and techniques to efficiently arrange convolution filters for efficient flow and integration of information between different feature map channels, achieving lower false detections. In this paper, we conduct detailed and rich experiments on a live-collected dataset and achieve 96.4% precision and 96.1% recall. The experimental results demonstrate that the CAC-Net model significantly reduces the misdetection and omission of fires and has efficient research value in the task of early fire detection in complex scenarios.

With the development of the network world, people's lives are gradually becoming intelligent. However, there are also crises in the network, such as attacks by some illegal hackers through a variety of complex means. The emergence of intrusion detection systems has greatly protected the network environment. However, traditional techniques cannot sufficiently learn the characteristics of attack traffic, resulting in poor performance. In order to address the limitations of traditional models, this paper proposes a transformer-based spatio-temporal feature learning method called STF-Transformer. It studies the temporal and spatial characteristics of the attack traffic through improved encoders and decoders, respectively. It greatly enhances the detection capability of the model. To demonstrate the effectiveness of the STF-transformer method. We measured the performance of the STF-transformer method on the CICIDS2017 and NSL-KDD datasets. The experimental results demonstrated that the STF-transformer method improved the detection rate. It achieved satisfactory results in terms of accuracy, F1 score, and other metrics.

Decreasing the correlation among the columns of the measurement matrix constitutes a significant technique for enhancing the quality of image reconstruction within the domain of compressed sensing. This paper introduces a novel one-dimensional homogenized composite mapping (1D-CICM) designed to generate chaotic measurement matrices for use in the compressed sensing process. The evaluation of information entropy, Lyapunov exponent, spectral entropy, and NIST SP800-22 test suggests that time series produced by the composite mapping possesses favorable non-periodic and uniform distribution characteristics, all while preserving its inherent chaotic attributes. The paper performed a correlation analysis on the measurement matrix constructed from time series produced by the composite mapping. The analysis shows that the columns of this measurement matrix have lower correlation among themselves. Further, the paper continues to construct an image compression-aware encryption algorithm centered on the 1D-CICM. By analyzing and evaluating PSNR, ciphertext pixel histogram, correlation, information entropy, and anti-differential attack test, the result analysis shows that the algorithm not only effectively improves the quality of reconstructed images at low compression ratios, but also improves the overall security of the encryption algorithm and its capability to withstand attacks.

The structure of the distribution network is complex, and the high-frequency signal of the traveling wave is susceptible to interference during the transmission process, which will lead to the lack of transient information and affect the acquisition and identification of the traveling wave characteristic data. To solve the problem that the fault transient traveling wave signal of the existing distribution network is disturbed by noise and leads to inaccurate positioning, a fault traveling wave positioning method based on Variational empirical Mode Decomposition (VMD), Sample Entropy (SE), and TEO is proposed. The VMD algorithm is used to decompose the fault-traveling wave signal into the Intrinsic Mode Function (IMF) component, and the low-frequency component of the noise is filtered out. The sample entropy algorithm is used to determine the noisy IMF, and the improved threshold method is used for processing. It is reconstructed and the Teager energy two operators are used to achieve accurate calibration of the fault traveling wave head. The experimental results show that the method proposed in this paper can have good denoising ability and accurately identify the traveling wave head of noisy faults. Compared with the traditional method, the denoising effect is better and the positioning error is smaller, which effectively improves the accuracy of traveling wave positioning in the distribution network.

Flexible grippers, characterized by their pliable "fingers," possess the ability to adaptively cover grasped objects, without prior knowledge of their exact shape and size. An example is the commonly used pneumatic gripper made of silicone material. These pliable grippers provide a better fit and protection for grasped objects, though they do impose limitations on the weight of the manipulated object. This paper formulates a composite material with combined mechanical and electrical properties based on Polyvinyl Alcohol (PVA) hydrogel. Designing a finger-like three-dimensional force sensor structure endowed with the capacity to enhance the hysteresis and repeatability of composite materials. The calibrated sensor is then affixed to the end of a robotic gripper, creating a semi-flexible gripper that combines the adaptability of a flexible gripper with the structural strength of a mechanical gripper. The algorithm proposed in this study, named TSO-BP, is based on improved BP neural networks inspired by tuna schooling behavior. This algorithm effectively fits the calibration data of the three-dimensional force sensor, establishing the mapping relationship between three-dimensional force and corresponding sensor signals. Additionally, the algorithm classifies signals generated during the semi-flexible gripper's manipulation of different objects. Experimental results demonstrate that this algorithm shortens convergence speed, enhances fitting precision, and achieves excellent fitting and classification accuracy.

Increasing the coherent accumulation time of signals is one of the effective ways to improve the long-distance detection performance of missile-borne radar for high-speed and large maneuvering targets. This paper closely combines the MTD process to study and propose a long-term accumulation method based on Radon transform under the quasi-continuous wave system. This method not only solves the distance and Doppler frequency movement problems caused by high-speed and large maneuvering targets, but also compared with other methods, it reduces the amount of calculation while maintaining efficient signal accumulation. Simulation results show that this method can effectively improve radar detection performance, is suitable for engineering implementation, and is of great significance to modern air combat radar detection technology.

The Global Positioning System (GPS) is widely used, which can provide users with all-weather and real-time navigation and positioning services. In a complex space environment, the propagation of GPS signals is easily disturbed by ionospheric scintillation, which results in signal strength decline. Aiming at the weak signal acquisition problem of GPS receiver, this paper proposes a weak GPS signal acquisition algorithm based on Double Block Zero Padding (DBZP) and improved differential coherent integration. After the signal is processed by DBZP, it is further processed by differential coherent integration, which effectively reduces the related power loss caused by Doppler frequency shift. In addition, the signal strength is enhanced by correlation accumulation technology, and the error caused by bit inversion of satellite signal is eliminated by bit inversion estimation technology. The simulation results show that the algorithm can successfully capture weak GPS signals, and the capture peak is more obvious, which improves the robustness of the algorithm. When the GPS signal suffers from severe ionospheric scintillation interference, the acquisition probability of Signal to Interference Ratio (SIR) still reaches 50% at -35dB. When the SIR is increased to -33dB, the probability will be further increased to 92%.

In order to overcome the interference caused by high-frequency noise and baseline drift when collecting pulse wave signals, this paper proposes a combination filter to preprocess the signal based on the traditional wavelet function denoising principle and threshold optimization rules. First, employ a digital low pass filtering technique to eliminate the power frequency disturbances present in the initial pulse wave signal.; second, improve the classical wavelet thresholding method and design an adjustable threshold function with flexibility and continuity; third, by establishing evaluation parameters, improved wavelet threshold algorithm for effect evaluation. Experimental results show that compared with traditional methods, the method proposed in this article has a good denoising effect on real pulse wave signals.

Navigation is one of the core parts of an AGV, which lets the AGV know where it is and guides its running direction and path. Aiming at the problems of poor map building and low map building efficiency of AGV in the navigation process, this paper studies the map building and positioning algorithm of AGV navigation system and verifies the effect of the optimized algorithm on the dataset. Two popular SLAM algorithms are analyzed theoretically and simulated, and Cartographer algorithm is chosen as the best one to incorporate the adaptive grid matching strategy into the algorithm. Theoretical analyses and experimental results show that the improved Cartographer algorithm improves the map building efficiency and map accuracy.

Frequency Diverse Array (FDA) provides pattern with range-angle-dependent beam by transmitting waveforms with tiny frequency difference. It is capable to contain both of the advantages by combining MIMO radar with FDA. At the transmit ends, the waveforms transmitted by the array elements are orthogonal with each other, meaning that the coherent integration is non-existent, and the jammer is hard to capture all the signals and transport them at the same time. At the receive ends, it is capable to get equivalent range-angle-dependent pattern through matched filters. Furthermore, adaptive beam forming algorithm is used to form null to restrain jamming. Simulations prove that the established system has effectiveness to suppress range deception jamming.

With the popularization of satellite communications and the rapid development of satellite network construction, the importance of satellite communications in military security and civilian fields is increasing. In this context, automatic identification of the modulation mode of satellite signals has become a key technology, which is of great significance in the fields of spectrum monitoring and spectrum management. In this paper, a deep learning-based signal modulation recognition method is proposed for the problem of low modulation recognition accuracy of satellite signals under the influence of Doppler shift. By taking the Doppler shift as the recognition feature, the CNN-GRU neural network model is used to recognize the signal. Verified by simulation experiments under different signal-to-noise ratios, the method achieves high accuracy in recognizing signal modulation modes such as BQSK, PQSK, 32QAM and 16QAM.

A sound vibration velocity measurement structure based on acoustic thermal field effects was studied. The measurement structure uses two parallel and suspended slender platinum wires, which are heated and cause proportional resistance changes under the disturbance of sound waves. The difference in resistance between the two platinum wires is detected to obtain the sound particle vibration velocity. The acoustic thermal field of the acoustic vibration velocity measurement structure was simulated and calculated using finite element method. New sensors are formed by integrating sound vibration velocity sensors and sound pressure transducer. The performance of the sensor was experimentally tested in an anechoic chamber. The experimental results indicate that the sound pressure-vibration velocity sensor has good frequency response characteristics, sensitivity characteristics, and directionality. Therefore, this sensor can be widely used in various near-field acoustic measurements, sound arrays, and sound source localization fields.

Radar Cross Section (RCS) is the most important amplitude characteristic of target electromagnetic scattering. Aiming at the low efficiency of low-resolution radar target recognition, a radar target recognition method based on one-dimensional convolutional neural network is proposed. According to the variation rule of RCS data sequence, the target RCS data sequence is generated by simulation first, and then the feature extraction of the target RCS data sequence is carried out by convolutional layer and pooling layer. Finally, the target RCS data sequence is recognized by full connection layer and Softmax layer. The experimental results show that the proposed method can classify and recognize the radar target and has good recognition performance.

The conventional multi-path coherent source moving target location method focuses on signal processing. Although it reduces the location interference, part of the sound source information may become invalid, affecting the location accuracy. Therefore, a joint localization algorithm based on DOA clustering algorithm for multi-path coherent source moving target and multiple sound sources is designed. Map the DOA sound source feature interval of the moving target of multipath coherent source, extract the DOA feature of the multi-path coherent source from the received signal, and provide basic conditions for subsequent sound source localization. Based on the DOA clustering algorithm, a joint localization model of multiple sound sources of source moving targets is constructed. With DOA characteristics as the basic condition, different sound source signals are separated through the clustering algorithm to ensure the accuracy of source moving target localization. Solve the optimal vector of multi sound source coherent source moving target joint positioning array, determine the optimal position of each sound source receiving point in the source moving array, and achieve multi sound source joint precise positioning. Simulation experiments show that the algorithm has higher positioning accuracy and can be applied in real life.

This article mainly obtains the position data and attitude information of anti-drone equipment through GPS/INS combined external guidance device, combined with target position data provided by active detection equipment such as radar, and applies Kalman filtering theory to preprocess external guidance data. After receiving target data, anti-drone equipment can judge and eliminate jump and garbled data and predict the current position of the target based on data transmission delay. At the same time, when the data volume is less than 10Hz, interpolation or smoothing can be performed on the target data to ensure the accuracy and continuity of the guidance information, and ultimately obtain the accurate position information of the target. Simulation and experimental results show that the external guidance of anti-drone equipment based on GPS/INS combination can achieve rapid search, recognition, and capture of unknown directional targets.

Signal decomposition is an important step in signal processing, and it is important to choose the appropriate decomposition method. In this paper, the principles and potential applications of EMD, NSP and VMD algorithms in various fields are introduced, and the priority order of signal components extracted by EMD, NSP and VMD algorithms is verified by experiments. The advantages and disadvantages of frequency domain resolution and decomposition of the three algorithms are compared. Among them, VMD has the strongest resolution in the frequency domain, followed by NSP. When it comes to breaking down FM signals, EMD and NSP are suitable, while VMD is not. Finally, this paper summarizes the advantages, disadvantages and application scenarios of these three algorithms, provides important guidance and reference for researchers to choose appropriate signal decomposition algorithms, and looks forward to future research.

Falls are one of the most serious health risk issues facing the elderly worldwide, which have a serious impact on the physical and mental health and quality of life of the elderly. Fall risk prediction can help develop targeted fall prevention programs that can help reduce the incidence of falls in the elderly. This paper proposed a fall risk prediction method based on electrostatic gait signals, measured the electrostatic gait signals of three types of people with different fall risks, and extracted gait time parameters, gait symmetry features, and gait variability features. The dimensionality of the dataset was reduced through the hybrid feature dimensionality reduction method based on the particle swarm optimization algorithm, and a fall risk prediction model was constructed based on the SVM algorithm, with the model accuracy reaching 96.77%. Methods of this paper have the advantages of simple equipment layout and non-invasive measurement, and can effectively predict the risk of falls, reduce the incidence of falls in the elderly, and improve the survival rate and quality of life of the elderly.

This paper primarily investigates a target localization method for passive radar systems using a multi-receiver technique. The study examines the direction of arrival (DOA) observation equations and the Least Squares solution algorithm, as well as the Time of Arrival (TOA) observation equations and the Gauss-Newton iterative solution principle. Based on the above research, a multi-station localization method that combines angle measurement with grid search for multi-surface intersection is proposed. This method addresses the issues of low localization accuracy in DOA method and the potential non-convergence of TOA iterative solution algorithm, achieving high-precision target localization.

Effective echo signal filtering is a waveform preprocessing step in airborne laser bathymetry. To address the low automation in effective echo filtering, we proposed an echo signal filtering method based on the K-means clustering algorithm. First, bathymetry signals were captured and saved as coarse signals using Matlab software. Subsequently, features such as maximum intensity, location of maximum intensity, and half-peak width were extracted from the coarse signals. Finally, K-means clustering algorithm was used to classify the samples, filtering out noise and abnormal pulse signals. The results show that K-means clustering is feasible for signal filtering, with noise signals filtering accuracy of 99% and abnormal pulse signals filtering accuracy of 95%, achieving standards comparable to manual filtering and significantly improving filtering efficiency. This method can further serve as a method for exploring preprocessing of a large amount of bathymetry signals.

Aiming at the problem that it is not easy to carry out water quality testing on the center waters such as inland rivers, lakes, large aquaculture ponds, etc., a remote water quality testing system based on the cloud platform is designed. The system is mounted on an unmanned boat. The microcontroller STM32 completes the control of the water quality testing cloud platform sampling, navigation system data transmission, and unmanned boat navigation. Among them, the water quality detection system can accomplish the detection of five kinds of substances closely related to water quality, including pH value, temperature, conductivity, TDS, and turbidity in water. At the same time, the microcontroller is configured with a wireless Bluetooth module. Through wireless transmission, the communication between the upper computer and the lower computer, transmission, and processing of test data is completed. Based on the LabVIEW platform, the Web client of the water quality testing system and the cell phone mobile remote monitoring system app were developed. It can display and monitor the test results in real-time and improve the efficiency and intelligence of water quality testing.

This paper presents an approach to the tow ship interference noise suppression based on the sector spatial prefiltering method. Firstly, the spatial spectrum estimation method is used to get the direction of the tow ship relative to the towed line array. Then, the spatial prefiltering method via sector inverse beamforming is applied to suppress the interference noise of the tow ship. Finally, the measured data are utilized to verify the suppression effect on the tow ship noise. The research results show that through this method, the signal in the passband sector could pass, while the interference noise outside the passband sector is suppressed, and the output data is still the data in the array domain after filtering. The experimental data analysis results show that this method has incomparable advantages compared with the beam domain method.

Urban rail transit is the backbone of public transport, which needs to maintain the daily operation order as far as possible. Due to aging equipment, technological innovation, unequal supply, demand, etc., renovation projects have become a key link in the life cycle management of urban rail transit. Signal system affects operation safety and efficiency, and it is necessary to avoid operation influence as much as possible. Given the potential risks in a renovation of signal systems have not been fully discussed, combined with a specific project of "CBTC + TBTC", the main risks related to the signal system are identified, the number of risk events and their levels are determined with the help of the risk grade matrix, and the risk management strategy is given accordingly, which has guiding value for the relevant engineering practice.

This paper mainly studies the architecture design and key technology of optical fiber communication digital twin system, so as to improve the performance and operation and maintenance efficiency of optical fiber communication system. Introduce the architecture design of digital twin system and explain the application of digital twin technology in optical fiber communication. Secondly, the key technology of optical fiber communication. Finally, we summarize the research work and discuss the direction of future digital twin system for fiber communication.

Under the influence of cold chain logistics technology, the cold storage industry is facing the challenges of planning, construction and scientific management while welcoming the opportunities of efficient, environmental protection and intelligent development. However, the traditional cold storage monitoring method has some problems, such as complex installation and wiring, high power consumption, difficult expansion and difficult maintenance, and it is difficult to meet the requirements of modern cold storage for operational efficiency and safety. In this regard, based on the internal environment of cold storage, this paper will integrate ZigBee technology, sensor technology and integrated development technology, and propose and implement a cold storage environment monitoring system based on ZigBee, so as to achieve the purpose of real-time, accurate and efficient monitoring of cold storage environment. Practice has proved that the system uses ZigBee network as communication medium, and the terminal node uses CC2530 as RF transceiver chip to complete the control of each sensor. The coordinator uses MSP430F149 single chip microcomputer as the control core, and cooperates with CC2530 to transmit data, which realizes the data interaction between the lower computer system and the upper computer system, improves the transmission power and receiving sensitivity, and at the same time increases the data transmission distance, expands the coverage of the cold storage environmental monitoring system, optimizes the monitoring operation process and enhances the system stability.

The transmission demand for deterministic delay, jitter, and bandwidth in emerging services is increasing. However, achieving common network transmission of various applications in the Wide-Area Network (WAN) while providing DiffServ poses a challenging problem. In this paper, our research team has devised a comprehensive deterministic network scheme based on SDN, which integrates FlexE, SRv6, CSQF, and other technologies to achieve deterministic assurance within the wide area network. Finally, we built a real network environment for testing. The test results indicate that after the calculation and configuration management of the SDN controller, the wide-area network can provide deterministic transmission and interference isolation.

With the continuous development of malicious hardware technology, HID attacks are becoming more and more capable and hidden. This paper provides a security protection scheme from the aspects of operating system login option configuration, including adding a CAPTCHA verification mechanism in the login process and an automatic lock screen setting. The scheme has been implemented and tested in a Windows 10 system based on the credential provider framework. The results show that the success rate of HID attack is close to zero in cooperation with well-educated users. Without user cooperation, the time between the user leaves the computer to the screen is locked is the only time that can be used by malicious HID devices, risking being discovered and cleared by users at other times. Compared with the existing schemes, this scheme has the characteristics of high security, low cost, and good compatibility. It can be implemented and deployed in different operating systems for computers with screens to resist HID attacks launched by a variety of malicious devices.

In recent years, the application of wireless communication technology has been developing from mobile smartphones to the automotive industry. The number of software configured on intelligent connected cars and the number of open network ports are increasing day by day, so they face more network security threats than traditional cars. The automotive software update framework not only needs to have high security and high performance, but also requires the framework to have customizable functions to meet the OTA (Over the Air upgrade) customization needs of car users. In order to improve the security of the framework while reducing the operating overhead of the framework on the vehicle side, and the framework has customizable functions, After analyzing and studying the Uptane framework, an optimization plan based on blockchain technology is proposed. Design a blockchain verification algorithm with low overhead for hierarchical verification and sub-regional verification, Improve the framework’s protection capabilities after key leaks.

In order to efficiently apply Wireless Power Transfer (WPT) technology in Photovoltaic (PV) systems, mathematical modeling and mechanism analysis of Magnetic Coupling Resonant Wireless Power Transfer (MCR-WPT) systems were conducted. By employing the multi-objective genetic algorithm (NSGA-II) and an adjustable coupling distance LCC-S topology, the reliability of power supply was ensured. The results demonstrate stable charging efficiency, strong adaptability, and high potential for widespread application.

Moving Target Detection (MTD) technology, as an effective means of suppressing clutter in radar signal processing, has been widely applied to various types of radars. In the development of traditional radar signal processing systems, the hardware computing resources of FPGA+DSP architecture limit the real-time processing performance of MTD, making it difficult for the radar to balance the clutter suppression effect and processing time well. In recent years, with the rapid development of GPUs, their powerful processing performance is gradually being applied to radar signal processing systems. In the process of implementing MTD, GPU relies on its powerful parallel processing ability to quickly suppress clutter and improve the radar's ability to detect moving targets. Starting from the design of the radar signal processing system, this article has completed the implementation of server-based MTD algorithm and the implementation of MTD algorithm in sliding window mode, improving the processing performance of the radar.

Structured Query Language Injection (SQLI) has always been a prominent threat in the field of cybersecurity, targeting web applications. Traditional techniques to prevent SQL injection typically rely on rule matching or dynamic and static analysis, which has high false positive rate and low detection efficiency. However, using deep learning algorithms can effectively detect SQL injection attacks and, consequently, overcome these issues. This paper proposes a SQL injection attack detection method based on TextCNN from the perspective of natural language processing and deep learning. Firstly, Word2Vec was used to process the text vectorization of SQL injection attack data, and then one-dimensional convolution was used to extract the features of the vectorized data. Finally, the trained SQL injection attack detection model was used to detect SQL injection attack. Experimental results show that the proposed method can effectively detect SQL injection attacks, and its accuracy rate, accuracy rate, recall rate, F1 value all achieved over 95%.

The rapid development of the Internet and the intelligent information age have provided more possibilities for data leakage, network attacks and other behaviors of malicious persons. The network security of party building in colleges and universities faces severe challenges. Intrusion detection is a new information security technology used to detect intrusions in computer network systems. Intrusion detection methods based on machine learning can better identify unknown anomalies than rule-based methods. The essence of the intrusion detection problem is a classification problem. There are many classification methods in machine learning, including convolutional neural networks, naive Bayes, SVM, etc. These algorithms have good classification performance on balanced data sets. How to extract more representative attribute features from complex data sets to more accurately describe normal behavior and intrusion behavior has become an important breakthrough point in current intrusion detection solutions based on machine learning. In order to solve the problem of network intrusion for party building in universities, this paper proposes to apply the convolutional neural network algorithm and random forest algorithm in machine learning to the intrusion detection model of party building network in universities, which effectively improves the overall classification accuracy of the model and the recognition rate of minority classes.

The automotive industry continues to develop, and the auto parts industry chain also needs to continue to upgrade. However, defects on the tubing surface can be a serious safety hazard, so defect detection is essential. Previous methods mainly rely on manual inspection, but due to the variety of surface defects, small area, small diameter and complex texture of tubing, manual inspection efficiency is low, and the accuracy is low. Faced with this challenge, the introduction of automated surface defect detection technology is imperative. This paper analyzes the characteristics and types of defects caused by automotive tubing in actual production and takes YOLOv8 as the baseline model to improve its feature extraction and fusion network part. The specific improvements are as follows: Add deformable convolution module (DCN) to the feature extraction network part of YOLOv8s. Flexibly extend the receptive field of convolution kernel to improve the accuracy of small target detection in low resolution images; The CA attention mechanism was added to the feature fusion network part of YOLOv8s. Strengthen the ability of the network to identify small target defects, so as to improve the accuracy of model detection.

This paper proposes a semi-supervised learning framework (SSL-SVM) based on Support Vector Machine (SVM) to address the complex features of modern data sets, such as high dimensionality and small sample sizes. The framework leverages unlabeled data through a self-training strategy, effectively addressing the issues of data scarcity and class imbalance, while improving the model's generalization ability, interpretability, and robustness. Experiments on liver cirrhosis datasets demonstrate that SSL-SVM is 14% more accurate than traditional SVM, providing a new strategy for data classification research.

Network security threats are increasingly becoming the focus of information system security, and an efficient and intelligent detection mechanism is urgently needed to deal with the evolving attack methods in time. This paper is devoted to exploring and evaluating network security threat detection algorithms based on machine learning. Through the empirical analysis of typical data sets such as KDD99, the performance of different algorithms in real network environment is compared. We have focused on algorithms such as Support Vector Machine (SVM) and Long Short-Term Memory (LSTM) in deep learning and attempted to explore their hybrid applications. Through comparative experiments, it was found that the LSTM+SVM hybrid model can effectively control the false alarm rate while improving accuracy. The experimental results on KDD99 data set show that our model has significantly improved the accuracy of many attack types, especially in the scene where the "U2R" attack type is "normal", the false alarm rate is about 9%. Through this study, I hope to provide valuable reference for the research and practical application in the field of network security threat detection and contribute to building a smarter and more robust network security defense system.