This PDF file contains the front matter associated with SPIE Proceedings Volume 12719, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

The aging of digital integrated circuits will have a certain [impact on the accuracy and reliability of circuits, and even pose a certain threat to the safety of circuits. To solve this problem, an IC aging prediction design method based on logic gate type is proposed. The aging characteristics of IC structure are identified by logic gate type technology, the aging degree index and evaluation algorithm of IC are standardized, and the aging prediction of IC is realized. Finally, experiments show that the IC aging prediction method based on logic gate type has high accuracy and reliability in practical application, and fully meets the research requirements.

In recent years, with the continuous development of the electric power industry in China, the per capita electricity consumption is increasing, and the scale of the power grid is also expanding. In order to ensure the power supply and continuity, a complete power supply system and power dispatching (PD) system must be established. At the same time, the key information privacy protection technology (PPT) of PD has also become the focus of the current power sector. Based on sensitive data (SD), this paper studies and analyzes the PPT of key information in PD. This paper discusses the key issues of SD security management and anti disclosure technology, as well as the advantages of SD security management and anti disclosure system model. The overall design of PD system is analyzed, and the PPT of key information of PD is studied based on SD. Finally, the effectiveness of PPT in this paper is verified through the evolution trend analysis of privacy leakage degree with the number of releases.

With the continuous promotion of the construction of "smart city", the service-oriented infrastructure such as hotels and hospitals are also beginning to realize networking and intelligent development. As an important part of the tourism industry, smart hotels are an inevitable way for the transformation, development and structural adjustment of China's hotel industry. The hotel needs to shift from the early extensive and homogeneous construction mode to the refined, intelligent and digital construction direction, and needs more distinctive products or services to establish its own brand advantages. In order to further promote the development of digital intelligence in the hotel industry. Through top-level architecture design, this paper dismantles the technical requirements of the smart hotel business, compares different networking schemes to help the hotel achieve high-speed, stable, good experience, easy management and full signal coverage of the intranet, besides, we analyse the 5G network deployment architecture model of the smart hotel. In addition, we also build solutions for three typical application scenarios: smart hotel AON (all-optical network), hotel operation management system and smart room control. At the end of the paper, the development of smart hotel technology and business is prospected, aiming to make a cutting-edge discussion for the development of smart hotels, and provide reference for the subsequent intelligent construction of smart hotel.

The traction substation is responsible for the power supply to the traction load, and it is very important to ensure the safe operation of the equipment in the traction substation. At present, the research on status estimation of primary equipment in the substation is mature, but there is still no unified evaluation standard and diagnostic process for secondary equipment. Therefore, an improved fuzzy algorithm model based on variable weight theory and membership correction is proposed to estimate the status of secondary equipment in the substation. The example analysis verifies that the improved fuzzy algorithm can improve the accuracy of state estimation and effectively avoid a series of problems caused by constant weights and fixed membership of the traditional fuzzy algorithm. The light transmission power decreases from -14.68 dBm to -19.74 dBm, while the temperature decreases from 23 °C to 28 °C. The health status of both indicators decreased by a level consistent with the observations.

With the development of electronic information technology and people's pursuit of a convenient life, intelligent and flexible wearable electronics have received more and more attention, and have achieved relevant research results. Conductive yarn, as the most basic flexible conductive textile material, the conductive yarn is widely used in intelligent wearable products. However, traditional textile materials are usually insulating materials and do not directly form conductive yarns and smart wearable products. This paper focuses on the preparation methods of conductive yarns by coating, spinning, and in situ polymerization, and the application research progress in wearable strain sensors, wearable supercapacitors and wearable electric heating devices. In addition, the convenience, comfort, washability and sensitivity required for wearable electronics are proposed to further optimise the flexible sensitivity and durability of conductive yarns and methods of combining textile materials with conductive materials are the main development directions for wearable flexible smart products.

With the development of the equipment maintenance, equipment health monitoring has become the focus of maintenance technology. Aiming at the spatiotemporal uncertainty of multi-factor coupling and failure in equipment health monitoring, a health monitoring system based on PLM using digital twin is proposed. Firstly, the system architecture and main functions are designed according to the characteristics of the health state of large complex equipment and the characteristics of digital twin technology. Secondly, the key technologies are proposed such as data converging, construct model, and prognose health in the digital twin monitoring system to ensure that the system can realize equipment health. Finally, a case study of TBM is proposed to verify the monitoring technology.

Heat flux of electronic equipment shows an increasing trend with the increasing demand for miniaturization and high performance. Low heat dissipation capacity makes it difficult to apply air cooling technology to high heat flux electronic devices in the future. At present, liquid cooling technology is the research hotspot of cooling solutions for electronics, mainly because liquid cooling has the advantages of strong fluidity and high heat transfer coefficient. Several kinds of liquid cooling technologies which can be used for cooling electronic equipment are introduced in this paper, among which micro-channel cooling, spray cooling, jet cooling and immersion cooling are emphasized. The working principle and progress in research of each liquid cooling technology are introduced and analyzed. Finally, the future development direction of liquid-based cooling technology is discussed. The problems in the research direction of liquid cooling technology are pointed out. Some suggestions for further research on liquid cooling technology are put forward.

Spectrum-induced polarization (SIP) is a widely used geophysical exploration approach, but it is prone to noise. To address this issue, this paper proposes a noise reduction method that combines phase space reconstruction (PSR) and fast independent component analysis (fastICA). The proposed approach enhances the signal-to-noise ratio of SIP data by effectively removing noise. Experimental results show that the PSR-fastICA method significantly reduces noise in SIP signals.

This article starts with the principle diagram of the road fork control circuit. By analyzing the working principle of the launching circuit and the working circuit, the common fault phenomenon of the control circuit is characterized by feature extraction. Finally, the common fault nodes of the ZD6-type rut-type cross-channel control circuit with dual -machine traction are given. When faults occur, how to efficiently and quickly locate specific fault positions, which can effectively guide on-site electrical personnel to solve the emergencies.

This paper mainly studies SFP56 optical transceiver, whose signal transmission rate is 50Gbps, transmission distance is 10km, working wavelength is 1310nm, working temperature is -40℃ to 85℃, and the small form-factor pluggables package structure is adopted. The SFP56 LR I-temp optical transceiver is mainly applied to 5G pretransmission networks to realize high-speed conversion between optical signal and electrical signal during optical signal transmission. The design adopts DSP technology, PAM4 encoding and TOSA scheme with built-in laser driver chip. The built-in laser driver chip is used to provide bias current and modulation current for the laser, thus realizing the stable transmission and reception of optical signals up to 53.125Gbps in 10km fiber. Compared with the traditional CDR technology and TOSA external laser driver chip scheme, DSP technology is based on digital signal processing. In addition to realizing clock and data recovery, DSP technology adds finite impulse response processing and multi-stage filtering in functions. Moreover, the scheme of TOSA built-in driver chip reduces the influence of impedance discontinuity and impedance mismatch of high-speed line between optical chip and electrical chip, so that the indicators of the optical module transmitter are better, the bit error rate of the receiving end is more stable, the power consumption is lower, and the reliability is better.

In view of the weak obstacle crossing ability and poor versatility of pipeline robots on the market, this paper designs an adaptive variable diameter pipeline inspection robot with obstacle crossing ability, which senses pressure changes through the pressure gyroscope and feeds back to the control center for feedback adjustment, and then the actuator variable diameter motor performs the diameter change, with the parallelogram reducer mechanism to enable the robot to stably change the diameter and walk, and the walking wheel mechanism is also equipped with a telescopic mechanism. Based on the principle of virtual work, the theoretical mechanical force analysis and ADAMS simulation of the mechanism were carried out to verify its rationality.

Aiming at the current problems of low-voltage distribution network (LVDN) at present due to the untimely updating of the account in the early stage, which leads to the unclear relationship between customers and affects the operation, management and fault research and judgment of distribution network, this paper presents an automatic identification method of distribution network topology on the basis of clustering of historical voltage data and its fluctuation characteristics. Firstly, the three-phase voltage data and user terminal voltage of the adjacent substation in the research object station area are collected through the distribution intelligent terminal, and the multi-dimensional fluctuation characteristic parameters of the voltage at both ends are defined and calculated, and the global fluctuation characteristics and local fluctuation characteristics of the voltage series are mainly analyzed; Then, the dimension of the target data is then reduced using the t ⁃ sne algorithm; Finally, based on the characteristics of voltage fluctuation after dimensionality reduction, the correct relationship between substation transformers and users was obtained by using an improved K ⁃ means clustering algorithm for clustering substation transformers and users. The result of the example shows that the subordinate relationship between the user and the transformer in the topology of the target station area can be identified through the detailed analysis of the voltage measurement data at both ends of the target station area.

In order to replenish the oil in the oil tank in time and reduce the loss caused by insufficient oil volume, we use wind turbines as an example to build a 1,300 oil tank picture dataset containing different oil levels. And the YOLOv5 network model based on the PyTorch framework trains the relevant dataset, and the oil tank and oil are detected through the training model to identify the oil volume of the oil tank, the model effect can meet industrial applications. The experimental results show that the YOLOv5 model established in this paper is 96% of the average accuracy of oil level recognition, which effectively solves the problem that the oil deficiency of oil tanks cannot be found in time.

In order to formulate classification adjustment methods for different abnormal power consumption reasons of smart grid, a strategy based on quadratic root function price adjustment and calculation demarcation is proposed, which is based on automatic process control and fuzzy comprehensive evaluation for monitoring, diagnosis and classification adjustment. Firstly, the exponential weighted moving average is used as the control variable of automatic process control and the upper and lower bounds are determined to monitor the power consumption, and the monitoring model is established; Secondly, the fuzzy comprehensive evaluation method is used to diagnose the cause of abnormal power consumption; Finally, according to different reasons, the classification adjustment strategy is formulated and the electricity price demand response adjustment model is established. The application of fuzzy comprehensive evaluation method can effectively analyze the causes of abnormal power consumption, and the development of classification adjustment strategy can eliminate abnormal power consumption and obtain stable fluctuations.

With the increasing number of energy internet users year by year, manual inspections are gradually being replaced by unmanned inspections. The target case algorithm based on mixed convolutional neural analysis has been widely applied in grid related snow, rain, ice, and wind intelligent cases. However, in practical applications, it has been found that due to the small size of the collected targets to be located, the accuracy of the mixed convolutional neural analysis model will decrease when the shooting angle is tilted and the lighting conditions are poor. This is because the hybrid convolutional neural analysis of this algorithm is relatively low. When there is a significant difference in angle or illumination from the case, the positioning and collection ability of the model will be severely affected. On this basis, the feature fusion part of ONNX algorithm and the selection of loss function and feature vector frame size are improved, and the improved CNN fusion method is used to classify various data in grid related snow, rain, ice and wind. Actual measurements and repeated experiments have shown that this method can be effectively applied to the recognition of various grid related snow, rain, ice, and wind data, optimizing mixed convolutional neural analysis, and greatly improving the recognition efficiency of grid related snow, rain, ice, and wind data.

The purpose of this paper is to extract the effective information of power grid operation by using various meteorological monitoring and forecasting information. In all meteorological disasters, considering that the damage to power facilities caused by strong wind accounts for 68% of the total disasters, and considering that the snow and ice disasters in 2008 caused serious losses in large areas, this paper makes prediction for meteorological disasters and derivative disasters such as strong wind, ice, snow in a certain area, studies the probability of grid failure caused by ice and estimates the severity of snow to estimate the probability of grid failure caused by ice. Finally, a decision support system for comprehensive prevention of power grid faults is established. The research and application of this system can scientifically arrange the planning and operation of power grid dispatching, and effectively realize the disaster prevention and reduction of power grid facilities and the refined management of power grid safe operation.

Distribution cable is an important part of distribution network, and its running state has an important influence on the normal operation of distribution network. At present, the maintenance of distribution cables is mainly based on offline regular maintenance mechanism, which cannot monitor the operation status of cables in real time, and the regular maintenance is not targeted and the monitoring method is relatively single. Therefore, the comprehensive evaluation method of improved fuzzy theory is used to evaluate the operation state of distribution cable, and a set of cable on-line monitoring device is designed. The designed device is installed and debugged on site, which verifies that the device can effectively monitor the state of cable.

Today's society has entered the era of industrial 4.0. LED power supply testing focuses more on efficiency and coordination of production and other factors. To this end, we invented the LED power wireless quality detection feedback device. This device takes the high-efficiency tester of LED power supply as the lower computer. Through the wireless communication interface JN5139 chip, the test results of the quality of LED power supply are sent to the PC computer of the upper computer, and the upper and lower current parameters of the lower computer can be changed by sending data through the wireless module in the upper computer. So our test system has the functions of online management and offline monitoring. Experiments show that the qualified LED power supply has good resistance to four important factors affecting surge current, such as lightning attenuation oscillation frequency, electrostatic discharge frequency, voltage sudden change frequency and switching action frequency.

The emergence of large and complex electronic equipment makes people more eager to improve its overall reliability and maintainability. For this reason, a comprehensive circuit detection system based on reverse simulation is designed in this paper. In this paper, according to the actual needs of the control circuit detection, the designed circuit detection system can simulate the sensor parameters needed in the testing process of the tested circuit board, effectively simulate the actual working conditions, and get rid of the dependence on physical sensor detection. The system can effectively solve the problem of ex-factory inspection of the control circuit board system, and is used for the maintenance and improvement of the circuit board to be tested. it is of great significance to save the cost, reduce the workload and improve the qualified rate of the products leaving the factory.

700MHz is known as the golden band, with wide coverage and strong diffraction ability, which is very suitable for deployment in vast rural and remote areas, and can greatly reduce the construction cost; With the proposal of the national "double carbon" goal, as an important field to achieve the "double carbon" goal, the State Grid of China has put forward new requirements for the construction of new power systems; The digitalization and intellectualization of the power industry put forward higher requirements for the delay, bandwidth and reliability of the communication system; The features of 5G technology, such as low latency, high reliability and Dalian connection, just meet the requirements of the above features; At present, the deployment planning of 700MHz network is still in the exploration stage. This paper combines the actual deployment and test of power 5G service, and provides a reference basis for a large number of deployment of 700MHz band terminals and base stations in the power system through the actual test data.

Wireless sensor technology can effectively sense the impact of the human body on wireless links, which is conducive to promoting the evolution of wireless networks to intelligent sensor networks. It has good application prospects in bright space, security, rescue, human-computer interaction, etc. Motion recognition based on wireless sensing has recently been a research hotspot. Passive RFID tags are battery-free, portable, and have good privacy protection. They can be transmitted in a non-line-of-sight range. Because of these advantages, this paper uses a deep convolution neural network to classify activities based on RFID Doppler shift data. We designed a complete yoga movement recognition and evaluation system to evaluate the whole design, taking 12 everyday basic yoga movements as examples. We designed and implemented the prototype of our scheme using an Inpinj R420 reader and RFID tag. We invited 30 volunteers to collect data and conduct comprehensive experiments. The experimental results show that the average recognition accuracy of the essential yoga activity recognition we designed is 93.20%. It can also be applied to recognizing other activities and has broad application prospects.

The traditional capacitor-less LDO is analyzed and a fast-transient response capacitor-less LDO with transient-current enhanced circuit is proposed in this paper. The overshot/undershot voltage can be detected when the load current makes changed, the transient-current enhanced circuit, based on AC couple network, can achieve high charging/discharging current of the gate of power transistor to improve the transient of the LDO. Based on the proposed frequency strategy, the on-chip compensation capacitor required is reduced and the area of the chip is minimized. The circuit is simulated in a 0.13μm standard CMOS and the quiescent current is 30 μA. When the load current changes from 1-50mA at 1μs, the maximum overshot (undershot) voltage is 79mV (46mV) Compared to the LDO without transient enhance circuit, it is improved 60.5% and 58.2%, respectively.

With the continuous development of intelligence, how to optimize the troubleshooting of power equipment is a subject worthy of study. The number of power equipment is large, and it is difficult to locate, troubleshoot and repair in time. Based on the research of power equipment troubleshooting based on 5G communication technology, a 5G communication acquisition module is designed. The data center platform uses multiple diagnosis strategies for diagnosis analysis and trend analysis, and uses mobile equipment as the terminal to make power equipment troubleshooting more convenient and intelligent.

This article develops a novel calibration method to calibrate automatic catchweighing instrument by studying the principle of the instrument, and using standard weights and test loads as calibration apparatus. The method evaluates the metrological performance of automatic catchweighing instrument by calibrating the weighing error in static operation, the weighing error and measurement repeatability in automatic operation, and the eccentricity both in static and automatic operation. The experimental results show that this method can effectively evaluate the metrological performance of automatic catchweighing instrument, and is of great significance to the quality control and metrological traceability of the instrument.

In order to reduce the response time and missed detection rate of 10kV distribution network power equipment operation status monitoring, a method for 10kV distribution network power equipment operation status monitoring based on digital twinning technology is proposed. Collecting and processing power equipment operation status signals using wireless sensors; Using digital twinning technology, virtual and real mapping is performed on the operation status of power equipment, and a twinning model for the operation status of power equipment is constructed; Using association rules to extract equipment operation status characteristics, the operation status monitoring of power equipment in the 10kV distribution network is realized. The experimental results show that the design method has a short monitoring response time and a low rate of missed detection, and has a good monitoring effect on the operating status of equipment.

Aiming at the limitation of factory testing of electronic water pump, the wind resistance load testing system of electronic water pump with gas as medium, combined with STM32 embedded system and various sensors is designed. The hardware part adopts STM32 as the main control chip, and designs the corresponding control circuit and signal acquisition circuit, and develops the corresponding upper computer software combined with VS2017 and QT framework, so as to realize the system automatic test. The system performs pump testing by applying reverse airflow, with fast testing rate and no additional pump cleaning work after the test is completed, which improves the efficiency of pump factory testing.

Modern radar has the characteristics of advanced technical system, multiple structural units, complex functional connections, and long continuous startup time, which requires high mission reliability. Fault prediction and health management technology can further improve the mission reliability of equipment while improving the inherent reliability of equipment through online monitoring of equipment status, real-time fault prediction and diagnosis, It is an important guarantee to improve the combat capability of equipment. Starting from the overall research requirements of modern radar, this paper demonstrates and analyzes the requirements of modern radar health management system, clarifies the system function composition, and proposes the system architecture scheme, with a view to providing a reference for the follow-up research on related technologies of health management system.

The radar equipment prognostic and health management (PHM) system evaluation index system is the foundation for the construction of the radar equipment health management system, and the construction of a scientific and reasonable index system helps to improve the quality of the construction of the health management system. This article takes radar equipment as the research object, combines the characteristics of radar equipment systems, uses tree analysis technology to analyze key issues, proposes the overall requirements for validation evaluation, constructs an initial index system for system evaluation, and studies the standardization of evaluation indicators and the classification of system evaluation levels, which can provide a methodological reference for the construction of radar equipment PHM system evaluation index system.

The reconfigurable multi-valued logic processor consists of a large number of processor units that can be grouped and reconfigured at runtime, enabling it to implement a wide variety of logic operators and many-bit parallel adders. In this study, we design and implement a demonstration system of an FPGA-based reconfigurable multi-valued logic processor using the AX7020 development board. To demonstrate the system's capabilities, we develop a demonstration program based on embedded Linux and Qt, which includes a graphical user interface for a 1000-trit reconfigurable ternary logic processor, a 32-quit reconfigurable quaternary logic processor and a 64-bit MSD adder. The program runs on the ARM Cortex-A9 processor on the PS side of the development board, with human-machine interaction achieved through the AN970 7-inch LCD touch screen external to the PL side of the development board. We verify the reliability of the reconfigurable multi-valued logic electronic processor's hardware functions in the demonstration system through various means, including software simulation checking, automatic random tests and manual tests on the demonstration system. The results show that the demonstration system can let users experience the characteristics of reconfigurable multi-valued logic electronic processor, such as processor units grouping and reconfiguration, ternary/quaternary logic operations and MSD addition operations, through the touch screen in a user-friendly way.

The accuracy of power load prediction has great influence on the actual power generation, distribution, system maintenance and the operation plan of energy supplier related to electricity price. This paper studies the application of feedforward deep neural network and recursive deep neural network in medium term power load forecasting and analyzes its accuracy and computing power. Firstly, a feature extraction method of time-frequency domain analysis is proposed to preprocess the collected original data set, which can fully excavate the deep information hidden in the original data set. Then the feedforward deep neural network and recursive deep neural network model are used to forecast the power load in the medium term. Finally, the actual load data of a city in 5 years are used to predict the load in different seasons in the next 1 year. The simulation results show that the collaborative use of time-frequency domain analysis method and deep neural network has higher accuracy in medium-term load prediction.

Aiming at the problem of power line target recognition in the process of power patrol inspection, this paper proposes a power line detection method that combines Hough and YOLO, as common detection algorithms often cannot accurately identify and determine the location of targets such as power lines and towers. Acquire image edge features through edge detection, and extract line features in the image using Houhg detection. After determining the line features to be selected, use the YOLO algorithm as a convolutional neural network framework to identify power lines and tower targets based on the migration learning process. Fusion of the target identification frame and the selected line is performed, mainly through rules such as intersection and slope judgment processes to eliminate interfering line segments, Finally, determine the exact location of the power line. After testing, the fusion method can well solve the problem of image line detection and location determination.

To improve the accuracy of geocoding system in processing offset in fuzzy address. This paper presents a method of offset processing in fuzzy address. First, obtain the offset azimuth and offset distance according to the offset in the fuzzy address, obtain the road information around the offset point. Select an appropriate road to offset according to the road direction, and obtain the returned address data.

Aiming at the problems of high loss and low satisfaction rates in the distribution process, this paper studies the distribution optimization of community group buying. Considering customer satisfaction, this paper studies the fixed cost of vehicles, the variable cost during transportation, the penalty cost for exceeding the time, the cost of goods damage during transportation, and the cost of self-pickup service in the process of distribution optimization. It constructs the optimization model of community group purchase fresh food logistics distribution considering customer satisfaction with the goal of minimizing the total distribution cost. According to the model, the genetic algorithm is used to solve it, and a numerical example is used to verify the conclusion. The optimization model of fresh food distribution in community group purchase considering customer satisfaction proposed in this paper is representative. This study can provide some theoretical reference for the planning of community group purchase distribution schemes.

This paper proposes an improved ASTER-based low-quality text image recognition algorithm to address the issue of blurry, low-resolution, and poor-quality text image recognition.The basic idea is to use super-resolution reconstruction technology EDSR to convert low-resolution text images into high-resolution text images, thereby enriching text feature information. In addition, the DCNV2 module is added to ASTER's feature extraction network to focus on the extraction of text feature information. Compared with previous methods, the method proposed in this paper significantly improves the recognition rate on seven public datasets, especially in the SVT and SVTP text image datasets, which have poor quality.

Aiming at the problem that the association query demand between the massive and complex relationship between flight test data in the flight test field cannot be realized at present, a knowledge graph construction method in the flight test field is proposed. Eight types of entities are defined and twelve types of relationships are designed according to the business requirements of association query, forming a complete set of flight test domain knowledge entities and relationship descriptions, and flight test data stored in the database are loaded into HugeGraph, a graph database, through data format alignment, entity schema construction, relationship schema construction and data extraction and transformation from the original database. Based on the attribute graph model, the flight test domain knowledge graph was constructed; a flight test data correlation query system was built based on the knowledge graph, providing a web query interface in graph form and displaying the query results to users through the D3 visualisation component with a graph structure. The experimental test of the flight test domain knowledge graph was conducted using the internal business system data as the source, and the results verified the completeness and reliability of the constructed knowledge graph, while the experimental verification of the constructed correlation query system showed that the query efficiency of the system was significantly improved compared with the traditional database method.

Face attribute editing is a research hotspot in the field of computer vision, which aims to modify a certain attribute of a face image to generate a new face image. The current methods based on Generative Adversarial Networks (GAN) have attribute entanglement problems and the implementation process is relatively complicated. To this end, this paper proposes a face attribute editing network based on style-content disentanglement and convolutional attention. Adding convolutional attention (CAT) module to the StyleGAN generator makes the network's control of content features no longer affected by the overall style of the image, and realizes the separation of spatial content and style from coarse to fine. In addition, the hierarchical CAT modules control different levels of attribute features, and changing the input of any layer of CAT can change the corresponding attribute features. The experimental results on the CelebA-HQ dataset show that the method in this paper can achieve disentangled editing of face attributes, and the scores of various indicators are better than the existing models.

Image segmentation is a typical task in the field of computer vision. Compared to the task of image segmentation within the framework of the hypothesis with a large set, recognition with a small data set can not only identify known categories in the training data set, but also assign colored labels to objects. In real-world segmentation tasks, due to various objective factors, it is usually difficult to collect exhaustive training samples for all categories when training a neural network. During training, there is a complete knowledge of categories about the world, and during testing, the algorithm is provided with information about the object (object mask). Segmentation by neural networks is required for accurate segmentation of existing categories in the training data set by assigning labels to them. In recent years, with the development of deep learning, deep learning for image segmentation relies heavily on various neural network models existing today. Although neural networks are very efficient, good results can rarely be explained. A recent method for solving the object segmentation problem involves two main stages: preparation of the initial images of a data set with specially prepared masks; and then training a compact neural network model using a specially prepared dataset. For an instance of a test object, when the probability is below the threshold value of a known class, it is determined that the object belongs to the category seen in the training data set. However, this method does not take into account information about the image data of unknown categories in the process of training the model.

Digital image correlation (DIC) technology has been widely used in engineering measurement. In the process of DIC strain measurement, the commonly used gradient method and the local least squares method for displacement field fitting often produce large errors. To reduce the strain measurement error and improve the measurement results, a method of local weighted regression smoothing and moving means smoothing for displacement gradient is proposed to measure strain. The simulated Gaussian speckle is generated by a computer, and the reference image is deformed to verify the effect of these two methods, then compared with the gradient and local least squares methods. The results show that the strain field obtained by smoothing the displacement gradient has a good denoising effect, a more accurate solution, a more stable result, and less error.

From the perspective of knowledge network, the research on the evolution law of technology network can understand the technology evolution mechanism and grasp the trend of technology development, which has certain guiding significance for formulating reasonable and effective technological innovation planning. Based on the PatSnap database, the patent data of the wind power industry (1900-2019) was obtained, the technical knowledge life cycle was divided according to the change of patent number and logistic s curve, and the knowledge graph of the patent network was constructed according to the citation relationship for analysis. The results show that the knowledge network evolves from the embryonic scalefree network to the small-world network, the connection between core patents weakens, overall patent node cohesion enhanced and the proportion of H02J3 classification technology expands.

Based on the scale phenomenon in nature, this paper proposes a scale algorithm. Under the guidance of the scale algorithm, the target image can be quickly collected. This paper analyzes the original shape of the image collected by the camera, image mosaic technology, image acquisition methods, compares the difference between the traditional target image acquisition method and the target image acquisition method based on the scale algorithm, and introduces the method application and implementation steps. Based on this method, relevant algorithm models and software can be developed to guide the practical application.

Facing the future far sea, high speed and wide range of far sea operations, based on the analysis and research of reliable distribution and timeliness requirements of data chain messages, combined with the high dynamics of satellite communication ground combat platforms causing dynamic changes in satellite network star-ground node topology, this paper proposes a multi-star multi-link channel access technology for data chain applications, which is based on real-time access to channel connectivity and routing information. This technique selects the optimal access platform based on the real-time acquisition of channel connectivity and routing information, taking into account the channel quality and time delay. The results show that this technique is suitable for efficient addressing of data link messages, improving the message delivery rate and reducing the network load.

Interrupt technology is a key technology for processors to respond to external events and plays an important role in the embedded field. The open source RISC-V architecture defines a complete set of interrupt mechanisms with hardware definitions that do not support interrupt nesting in embedded systems that only support machine mode, and software support for interrupt nesting adds additional time overhead. This paper is based on the interrupt mechanism defined by RISC-V, using a state machine to control the global interrupt enablement of the processor to support interrupt nesting, and to save the site and restore the site through hardware stacking and out-stacking to avoid the time consumption caused by the software level; a dedicated entry point for interrupt service program jumps further speeds up the response. In this paper, we design and implement a fast interrupt system using Hummingbird E203, an open source kernel defined by standard RISC-V, as an experimental platform. Simulation verification and comprehensive implementation on FPGA platform show that compared to the interrupt handling process of Hummingbird E203, using a small amount of hardware resource consumption, the response speed is improved by 1/4, providing better real-time performance and flexibility in embedded applications.

Operation ticket issuing is one of the daily work of substation operation and maintenance. The current traditional invoicing mode has some problems, such as insufficient intelligence and efficiency. In order to improve the existing invoicing mode, shorten the draft time, reduce the error rate, and improve the quality and efficiency, this paper proposes an intelligent invoicing and management system based on Robotic Process Automation (RPA) technology. RPA robot is used to replace traditional manual invoicing, realizing automatic batch invoicing of the whole process of corresponding operation tickets, and simplifying the mechanized invoicing process of operation and maintenance personnel. Through the management system, the completed operation ticket is photographed and uploaded, the training set is built using the neural network model for character recognition, and the corresponding operation ticket content is backfilled to achieve the statistical analysis of the operation ticket data. Practice has proved that the input of the system can effectively shorten the invoicing time, save resources and improve work efficiency.

Blood vessels are the main anatomical structure of the fundus retina. Retinal blood vessel segmentation images have been widely used in the judgment of cardiovascular and cerebrovascular diseases and retinal diseases. Therefore, appropriate fundus retinal blood vessel segmentation method is of great significance for the detection of retinal diseases. Based on U-Net, the original convolution structure in the encoding part is replaced by the Res-Se module, and the CBAM module is introduced in the skip connection part to achieve fine-grained feature fusion, thereby improving the network's ability to segment the subtle features of retinal vessels. Experiments on the CHASEDB1 dataset show that the proposed model has certain improvements in accuracy, sensitivity, and specificity indicators. This model can more accurately segment retinal vessels and demonstrate better segmentation performance.

In order to solve the problem that the Yin-Yang Pair Algorithm (YYPO) tends to converge prematurely and has low convergence accuracy, a Yin-Yang Pair Optimization Algorithm (CYYPO) that combines Cubic chaotic mapping and lens imaging learning strategies is proposed. The algorithm uses Cubic chaotic mapping to initialize the two points of yin and yang, and uses the strategy of lens imaging learning method to design a new forward and reverse search segmentation method, which is tested by CEC2013 standard test function, and the results show that compared with the most basic YYPO and other five intelligent optimization algorithms, CYYPO has higher computational accuracy and more outstanding performance advantages.

This article aims to address the issues of high cost, poor stability, and susceptibility to external interference in centralized wiring for refrigerators. A low-cost and highly stable distributed bus architecture based on customized LIN bus technology is proposed to replace the traditional centralized wiring system for refrigerators. This bus architecture supports up to 64 nodes and allows flexible addition and removal of nodes, while also featuring redundant short-circuit failure protection and anti-bus noise interference capabilities. This article provides a detailed description of the data transmission system based on STM32 and various security mechanisms, such as communication ID verification and checksum verification, to ensure communication security and stability. The experimental results demonstrate the feasibility of the proposed solution, which has already been applied in actual project development.

In order to classify glass artifacts into different categories based on two characteristics: the chemical composition of the sampling points and whether they are weathered or not, this paper is divided into two parts. In the first part, the entire dataset is divided using data preprocessing, and a logistic regression model is constructed for binary classification. The optimal parameters of the model are estimated using maximum likelihood estimation and gradient descent algorithm. The aim is to explore the classification patterns of the two types of glass artifacts. In the second part, unsupervised learning k-means algorithm is used for clustering. Indicators are selected based on mean square error and confidence level. The same model as in the first part is used to test the effectiveness of sub-classification. The results show that glass artifacts can be divided into two main categories: high-potassium glass and lead-barium glass, and further subdivided into 14 subcategories.

Since deep learning has been continuously developing, research on language models utilizing deep learning has increased, and certain results have been obtained. There are many research results, but RNN has unique advantages regarding the construction of language models. In addition, due to its weight-sharing cyclic network form, the language model can have the information memory function described above, which facilitates the prediction of content in the future. Text generation is one of the most common applications in this area. However, RNN may have some limitations, including its inability to remember the above for a prolonged period of time, making the language model formed by RNN relatively weak in terms of intelligence and ability to connect the languages under study. Therefore, this paper proposes to build a language prediction model based on the RNN network model LSTM. This will allow the language model to capture the previous characteristics better, resulting in improved intelligent text prediction.

This paper design of a time resource allocation management method based on TDMA systems. The method integrates and manages the time resources of all dedicated access modes in the whole network, designs a dynamic adjustment algorithm to achieve dynamic allocation and recovery of resources, and considers and analyses the possible exceptions. The application of the method makes the overall network highly flexible and operable, as well as scalable.

In the traditional visual relationship detection model, the entity representation of visual features tends to focus on coarsegrained and ignore fine-grained features. Moreover, the entity representation of spatial features does not fully reflect the prominent role of relative position. In some specific cases, it is impossible to generate a unique spatial feature representation vector. The final feature fusion did not take into account the characteristic that visual features are primary, while semantic and spatial features are secondary. In order to address the above issues, we propose a visual relationship detection model with pixel-level feature enhancement and weighted fusion. Specifically, we embed a fine-grained information block to capture pixel-level context information in the feature map, providing richer visual features for the prediction of relationships. We adopt a coordinate encoding method to encode the respective and relative positions of entity pairs to bounding boxes to obtain more accurate spatial feature representations. We construct a feature fusion method based on feature weighting to obtain a more improved fusion vector. We conducted extensive experiments on the mainstream visual relationship detection dataset. The results show that our proposed model performs better.

Internet of Things (IoT) data is characterized by large amount of data and high real-time performance. Processing IoT data through complex event processing techniques requires setting complex rules, but the rules tend to change as the business changes. Rule engine can set rules through separate configuration files to filter to data or things matching the rules without modifying device data or management platform code. In this paper, in order to solve the performance of the rule engine at large data volume, we design a rule network-based reduced-dimensional K-Means raw data clustering method to cluster devices by the difference of device operation data while implementing parallelized rule networks according to the rule types. After the experimental analysis of the method, the results show that the reduced-dimensional K-Means raw data clustering method based on rule networks has smaller event loss and effectively reduces redundancy.

Fog computing is a cutting-edge distributed computing architecture that optimizes resource allocation by fully utilizing the resources within the computing network. However, data security is a major concern in fog computing. The implementation of existing Ciphertext-Policy Attribute-based Encryption (CP-ABE) schemes in the fog computing environment faces challenges such as communication blockage, insufficient security, and poor performance. To tackle these challenges, this paper introduces a novel Fog-assisted Revocable and Auditable Access Control System (FRAACS), which incorporates fog-assisted key generation and decryption, user revocation, and auditing mechanisms. The comparative analysis of the proposed scheme's functionality and performance shows that it is a viable, secure, and fitting scheme for the fog computing environment.

In the object detection, the processing of feature fusion and the structure of detection head have an important impact on the detection performance. The current detectors often use the detection pipeline of ‘backbone-feature fusion network-head’. We first propose the Feature Fusion (FF), which constructs a lightweight branching network based on the image pyramid and fuses its extracted features with those of the backbone network, providing a new idea for the focus of feature fusion. In addition, we design the Shared Detector Head (SDH). The main purpose of SDH is to reduce the inconsistency of predictions on feature maps between classification and regression tasks, enhance the interaction between the two, and enhance the detection performance. Our experiments on MS COCO2017 and PASCAL VOC0712 datasets support the above analysis. Based on the above improvements, our approach achieves 0.8% mAP improvements on MS COCO2017. The above experiments prove the effectiveness of our approach.

Unplanned construction operation on the site of power transmission and transformation project occurs frequently. In order to catch up with or speed up the construction progress, the construction site does not report the plan and arranges the operation without permission. This kind of unplanned operation lacks monitoring and is untracked, which inevitably leads to inadequate safety measures, leading to greater construction safety risks. Satellite remote sensing technology has developed rapidly. It has significant advantages such as wide coverage, a stable renewal cycle, less limited by environmental conditions, and a large amount of information. At the same time, it is completely independent of the construction group. It can provide objective and full coverage visualization of the transmission line construction project. It is an effective technical means to carry out unplanned operation investigations and random inspections.

The emergence of BERT and Knowledge Graph (KG) has promoted the development of Question Answering (QA), however, existing QA systems are still inadequate in terms of the accuracy of relational reasoning and the interpretability of answers. In this paper, we combine BERT and KG, following and optimizing existing methods to build a medical QA system with better performance - KBMQA. The final experimental results show that KBMQA performs better on both MedQA and MedNLI datasets compared with previous biomedical baseline models and MOP models.

The Message Queuing Telemetry Transport (MQTT) protocol is commonly used in the Internet of Things (IoT) due to its simple and efficient communication method. However, this also brings some security risks, such as arbitrary device publishing and subscription that can cause the IoT system to malfunction. Traditional MQTT access control methods, such as using access control lists (ACL), have some problems, such as security issues in storing permissions and matching efficiency. In this paper, we propose a blockchain-based access control method for IoT MQTT communication, which uses decentralized blockchain to store access control information, solves the problem of single-point failure, and ensures information security. To address the MQTT permission matching issue, we propose a Patricia Trie based adaptive priority topic search tree (PTST). This data structure optimizes the storage consumption of permission information and, for MQTT wildcard matching, uses an adaptive priority permission matching approach, which always searches from high to low priority, greatly improving the matching efficiency. Then we propose an access control model and process based on eXtensible Access Control Markup Language (XACML) standard. In the end, we perform performance testing and comparison with other data structures, and the results show good performance of PTST, demonstrating the feasibility of our scheme.

Many studies have shown that wireless sensing would be a promising method for liquid identification, but existing methods still have limitations for fine-grained liquid sensing. In this paper, we propose a liquid identification method based on multiple transceiver pairs, which effectively improves the sensing resolution of liquid identification. We implement our method with a commercially FMCW millimeter wave radar and evaluate its performance. Our result shows that for concentrations as low as 0.5% in alcohol solutions, our method can achieve an accuracy of more than 95%.

Through hash learning, the image retrieval based on deep hash algorithm encodes the image into a fixed length hash code for fast retrieval and matching. However, previous deep hash retrieval models based on convolutional neural networks extract local information of the image using pooling and convolution technology, which requires deeper networks to obtain long distance dependency, leading to high complexity and computation. In this paper, we propose a visual Transformer model based on self-attention to learn long dependencies of images and enhance the extraction ability of image features. Furthermore, a loss function with multiple loss fusion is proposed, which combines hash contrastive loss, classification loss, and quantization loss, to fully utilize image label information to improve the quality of hash coding by learning more potential semantic information. Experimental results demonstrate the superior performance of the proposed method over multiple classical deep hash retrieval methods based on CNN and two transformer-based hash retrieval methods, on two different datasets and different lengths of hash code.

The research of Internet of Vessels high-precision navigation positioning integrated equipment analyzes the integration problem of ship navigation equipment system, fuses and processes AIS and radar ship travel trajectory information to enable ships to travel safely in complex maritime environment, and solves the problems of low navigation positioning accuracy and limited detection capability. Kolmogorov-Smirnov test, Gauss-Kruger projection and other methods are used to solve the problems of outlier removal, coordinate unification and distance correlation judgment derived from the fusion of AIS and radar information. The simulation results of the ship travel trajectory fusion algorithm for AIS and radar indicate that the radar and AIS information can be highly fused, solving the problem of large errors in the process of radar and AIS target trajectory fusion.

The recurrent neural network model based on attention mechanism has achieved good results in the text summarization generation task, but such models have problems such as insufficient parallelism and exposure bias. In order to solve the above problems, this paper proposes a two-stage Chinese text summarization generation method based on Transformer and temporal convolutional network. The first stage uses a summary generation model that fuses Transformer and a temporal convolutional network, and generates multiple candidate summaries through beam search at the decoding end. In the second stage, contrastive learning is introduced, and the candidate summaries are sorted and scored using the Roberta model to select the final summary. Through experiments on the Chinese short text summarization dataset LCSTS, ROUGE was used as the evaluation method to verify the effectiveness of the proposed method on Chinese text summarization.

In this work, we propose a new algorithm framework based on Transformer to address the distortion problem of scene text images caused by shooting angles or lighting conditions. This new algorithm framework is called Scene text image transformer (StTr). Specifically, we use a feature extractor containing six convolution modules for feature extraction. Then, we flatten the obtained feature maps and input them into the Transformer encoder and decoder. After obtaining the displacement matrix, we use it to perform geometric correction on the input image. We also use Transformer to eliminate the lighting problem and improve the accuracy of OCR recognition. We evaluated our StTr on different datasets and achieved a 16.73% improvement in accuracy compared to the state-of-the-art methods. Additionally, our model has the advantages of being lightweight and efficient.

In order to remove noise effectively, a new wavelet threshold denoising algorithm is constructed on the basis of traditional wavelet theory. First of all, select appropriate wavelet basis function and and decomposition level, and then appropriate threshold function and threshold value are selected for threshold processing. At last, the denoised image is derived by reassembling both the low-frequency and high-frequency components. The adjustment parameter is introduced into the threshold function to make it flexible in denoising. Through Matlab simulation, the evaluation index results of several denoising algorithms are compared. The experimental results show that the subjective visual effect and objective evaluation index of this algorithm is obviously better than other algorithms, and it has a certain level of practicality.

Bionic polarized light navigation technology is an emerging autonomous navigation technology with the advantages of full autonomy, anti-interference, and non-accumulation of errors. This paper reviews the research progress of bionic polarized light navigation technology from three aspects: research on sky light polarization distribution patterns, polarized light navigation sensor technology and combined navigation methods. This paper firstly compares the research progress on the sky polarization pattern in recent years, and introduces the atmospheric polarization pattern analysis model and underwater polarization pattern analysis model; reviews the research status of polarized light navigation sensors at home and abroad in recent years, and divides the sampling mechanism of sensors into point-source and imaging polarized navigation sensors and introduces them respectively; the current development of polarization sensors towards miniaturization and integration makes it possible to integrate polarization sensors with other navigation systems, and a combined navigation method with polarized light navigation sensors is introduced. Finally, the development trend of bionic polarized light navigation technology based on the polarization distribution characteristics of sky light is discussed, and the key issues to be solved in the development of navigation technology are presented.

The search function of shopping software is essential, and good recommendations can increase the user's desire to buy. The fuzzy search and image similarity search proposed in this paper is a new retrieval method built on Elasticsearch, which can speed up the search and improve the retrieval's correctness. Its support for various complex texts dramatically facilitates the development of this project. This search type is used in home shopping software to improve the user's comfort significantly. The text is developed and designed based on the Golang language, whose high concurrency and excellent library functions help implement the functionality extensively. The user side is presented as a WeChat applet, which lowers the threshold of use and increases the dependency of users. With Elasticsearch's support for multiple languages and its unique vector search and text embedding features, the system can train models such as Contrastive Language-Image Pretraining (CLIP) and Natural Language Processing (NLP) on different images and languages, improving the search's accuracy. For the model generated by the training, vector search is performed to achieve the purpose of the search, and finally, the search results are returned to the front-end applet page for exhibition.

An important part of the dialogue robot is the understanding of semantics and the analysis of sentences. At present, most of the models are trained based on character vector and word vector. Compared with traditional models, this method can fully consider the long and irregular language characteristics of daily chit-chat dialogues, and make full use of character and word fusion to improve the dialogue performance of the model. At the same time, the word embedding matrix trained by the pre-trained model is calculated to obtain the same dimension as the character embedding. Then, the character vector and word vector are fused, and the transformer model is used for dialogue training. The experimental results show that the proposed method can better improve the understanding ability of the dialogue model.

In order to improve the feature extraction process in the semantic segmentation network and clarify the operating logic of the model, this paper splits the discriminative features according to the pixel's own information and the inter-pixel correlation information. Using two branch networks, relational branch and information branch to focus on the extraction of their own features, and using the fully convolutional segmentation head and the fully-connected segmentation head respectively to make the features more in line with the requirements of the branch itself. In order to make full use of the information contained in the discrepancy between the two branches, a Discrepancy-guided Fusion module is proposed at the feature level, in where the difference information between the features is used as a guide to promote the fusion of the features of the two branches. At the prediction level, with the help of Point-wise Discrepancy Cross Entropy loss, the prediction difference is used to determine the network's attention to each pixel. The effectiveness of the method is verified by extensive experiments on the Cityscapes dataset.

Based on the structural characteristics of semi-trailer trains and the problem of large blind spots in their field of view, this article proposes a pure visual solution to determine the articulation angle technology and build a panoramic view system for semi-trailer trains to solve the problem of blind spots. Firstly, the fisheye camera is corrected for distortion and perspective transformation, and a weighted fusion algorithm is used to generate partial panoramic images of a single vehicle body. Then, the articulation angle algorithm based on feature point matching is used to obtain the articulation angle information, and the panoramic view image of the entire vehicle is generated by combining the articulation angle information. Finally, a simulation scene is built in Prescan to verify the stitching effect. The simulation results show that the proposed panoramic view system can effectively and clearly generate panoramic view images of semi-trailer trains at different articulation angles, effectively eliminating blind spots in their field of view and improving vehicle safety.

With the explosive growth of the number, type and quantity of terminal access, 5G networks need to support a large number of diversified business scenarios from the vertical industry to meet the different needs of differentiated services. Network slicing technology is the key to realize the vision of all things interconnection with wide connection, high reliability and ultra-low delay services in 5G. How to unify different network slicing according to the business needs of different scenarios has become an important issue in the implementation of network slicing. This paper first studies the current development status and technical hotspots of 5G network slicing, and then designs the network slicing schemes for different scenarios. Finally designs a network slicing scheme in detail for multi-service in eMBB scenario, and the feasibility of the scheme in practical application is verified Via XLABS simulation platform.

External radar has obvious advantages in “low, slow and small” target detection. With the large-scale construction of 5G civil communication network, 5G communication signal will become the richest electromagnetic resource in the future battlefield environment. Therefore, it is of great significance to study the positioning method based on 5G external radiation source. In this paper, the distance sum positioning method is applied to the detection of 5G external radiation radar, and the distance sum positioning model of 5G external radiation radar is established. The influence of base station distribution and distance sum error on positioning accuracy is simulated and analyzed. Based on this, some useful conclusions are given to improve the positioning accuracy of 5G external radiation radar.

When using UAVs to inventory goods in large warehouses, multiple targets need to be inventoried and the goods dropped off. Precise positioning of the UAV is necessary for accurate inventory tasks. The optical flow localization scheme is only suitable for the environment with good lighting conditions and rich ground texture. On the other hand, Lidar or UWB requires a higher cost. In this paper, we use optical flow and vision to achieve precise positioning of multitasking nodes. Optimizing the path of drones using the Floyd algorithm. During flight, vision is used to correct the accumulated error generated by the optical flow. We have carried out simulations and experiments. The result show that our scheme is effective and feasible and has the advantages of low cost and high accuracy.

Long short-term memory (LSTM) is a widely used artificial neural network that is well suited for time series prediction. Quantum machine learning as a new research topic combines the advantages of quantum data processing and classical machine learning. In this paper, based on a hybrid quantum classical scheme, we design a quantum enhanced LSTM model and several variants such as QGRU. We also performed experiments with a multivariate time series prediction problem to verify the feasibility of these models. Through this research, we expect to explore the benefits and implementation of quantum-based machine learning.

The life of most industrial instruments is closely related to the local climate, and there is a significant correlation between the development of industry and the local climate. In order to explore the future global temperature distribution pattern and temperature level under the global warming trend, the global temperature data published in Worldclim is used. The XGBoost-RedRNN++ model is proposed in this paper to forecast the future temperature, and the regional global temperature prediction is carried out with 5 'as the precision unit. After predicting global average temperature of 20.78 centigrade in 2050 and 22.54 centigrade in 2100. The similarity of the global average temperature distribution in 2050 and the average temperature distribution in 2021 is 95.54%, and the similarity of the global average temperature distribution in 2100 compared with 2021 is 81.21%. With time going on, the temperature is generally rising everywhere. The global lowest temperature in 2050 and 2100 will be in Antarctica, with an average annual temperature of -14.6 centigrade and -12.4 centigrade. The highest global temperature in 2050 and 2100 will occur in Central Africa, with an average annual temperature of 85.2 centigrade and 87.5 centigrade. Additionally, it was observed that the temperature gradually decreased from the equator to the poles, and the land temperature in the southern hemisphere was significantly higher than that in the northern hemisphere. This study can not only lay the foundation for the future industrial development, but also provide theoretical support for the reform direction of the national industrial constitution.

This paper evaluates the container ports importance along "21st Century Maritime Silk Road". A directed weighted container shipping network is constructed by collecting global container ship route data in 2021 and using complex network related methods. This paper proposes a comprehensive evaluation model with the Entropy-weight TOPSIS method based on the ports’ own multi attributes involving the navigation, transfer and anti-interference ability. It also proposes an evaluation model based on the ports’ global attribute considering the importance influence of other ports who has communication with the evaluated port in the network. 400 container ports’ importance in the real container shipping network along the "21st Century Maritime Silk Road " are evaluated by using the above two evaluation methods and the results are analyzed to prove the correlation and rationality of the evaluation methods. The result shows that Singapore Port, Rotterdam Port, Kelang Port, Hong Kong Port, Ningbo Port and Shanghai Port are very important in the container shipping network; The important ports identified based on the ports’ global attribute evaluation method have greater influence on the container shipping network, and the port failure will lead to more serious network damage; In order to ensure the stability of the container shipping network, the key ports which has high ranking in importance evaluation should be given priority to development and protection. In addition, strengthening the joint construction of the key ports themselves and the surrounding port groups is also significant.

In the computer vision field, vehicle re-identification research is an important direction. There have been many researchers doing research on vehicle re-identification. But they mostly focus on the overall image and ignore occluded images, which may be more practical and challenging. This paper proposed a re-identification method to solve the occlusion in remote sensing image, which extracted the pixels in the random rectangular area of the image and covers them to other positions of the image. The feasibility of the proposed random occlusion method was proved by experiments on simulated occlusion images. And experiments on real occlusion image data sets showed that the proposed random occlusion method was better than the random erasure method in solving the occlusion problem, and can obtain better re-identification results.

Bullet screen is a new type of comment appearing on video websites. When the video is played, it will be displayed in real time. Through sentimental analysis of the screen, we can quickly understand the sentimental types of the current video. However, it is unrealistic to deal with all the bullet screens of the whole video at one time with traditional methods. This paper proposes a method of aggregating all text content in video to represent very long video content to achieve video feature extraction. The method proposed in this paper is more effective than PCA and scalable methods in reducing dimension and achieving aggregation.

As a new technology, UWB technology has great development prospect. In this paper, the basic principle of UWB technology is briefly described, several common UWB technologies are emphatically expounded, and the characteristics of each technology and their advantages and disadvantages are compared, and the development trend of UWB technology in the future is proposed.

High-speed coherent optical communication systems can effectively solve the current problems of transmission capacity and transmission distance in the field of communication, and have become an important research direction in optical communication. The optical signals generated by lasers are composed of different frequency components, which have different transmission speeds in optical fibers, causing dispersion effects. Dispersion causes inter-code crosstalk, which has an impact on coherent demodulation and threshold judgments at the receiving end. Dispersion has a constraining effect on the performance of high-speed coherent optical communication systems, and the development of optical communication requires the study of equalization methods to solve the dispersion problem. In this paper, we examine the dispersion problems and the corresponding solutions in coherent optical communication by reviewing various domestic and foreign literature, focusing on the classical optical communication dispersion equalization technique and analyzing the phenomenon of limited dispersion equalization effect due to the constant modal value of the classical technique tapping coefficient. We study and design the dispersion equalizer CMGD-FIR based on the momentum gradient descent optimization algorithm, which exponentially weights the average optimized gradient and corrects the deviation at each update iteration. Simulation results show that this technique can quickly update the tap weights close to the minimum value and improve the convergence speed while ensuring the dispersion equalization effect.

With the advent of the information age and the continuous development of communication technology, the power communication system is also being updated and upgraded. The transmission speed and quality have been greatly improved. Its development direction is high-speed transmission, which can effectively drive the development of other industries. This development trend has led to the emergence of 5G mobile communication technology, which has been tried and applied in the power communication system, and has now played an important and positive role. This paper expounds the advantages and characteristics of 5G mobile communication technology, and analyzes the specific application of 5G mobile communication technology in power system communication from different aspects in combination with the actual situation, with a view to providing reference.

Monocular 6D pose estimation is a fundamental task in computer vision, and this paper focuses on class-level 6D pose estimation that can predict the pose of previously unknown objects. In the work carried out on the basis of RGB-D images, previous approaches have paid less attention to the distinction between different structural features in them and the consistency of their contributions in the pose estimation task when using deep learning for feature extraction, and thus most of the fusion is a direct stitching of heterogeneous information features. In particular, the experimental results of the current work on 6D pose estimation surface that there are still limitations in its pose estimation for multi-class target training. The sparse structure of the point cloud would make most methods easy to ignore the effective features. Moreover, the existing method studies do not sufficiently explore the complementarity between the effective information of heterogeneous features, which can lead to the fusion of the optimal combination methods that lack the contribution of each, thus bringing a large amount of redundant information and consuming computational resources. Therefore, this paper designs a more effective point cloud feature extraction method for dynamic graph structure for this task. To address the inherent requirements of complementary fusion, we design an adaptive method for further feature extraction on heterogeneous data and a fusion method with effective self-attentive disequilibrium contributions to extract core information from potential feature complements quickly, accurately, and efficiently. We conducted experiments on popular benchmark datasets, such as the NOCS-REAL [1] dataset. The experimental results show that our proposed method can perform the multi-class target 6D pose estimation task end-to-end and has a good performance on these datasets while achieving a real-time inference speed of almost 20 FPS.

Unmanned Aerial Vehicle (UAV)-powered Wireless Sensor Networks (WSNs) are considered to be a promising solution to the problem of the limited power of the Sensor Nodes (SNs). In this paper, we introduce a UAV-powered WSN system, where multi-UAVs undertake the role of remote charging stations. To optimize the overall power efficiency, we design the collaborative trajectories of the UAVs. In order to solve the problem of trajectory planning, we first model the service process as a Markov decision process (MDP), and then propose a Multi-Agent Deep Reinforcement Learning (MADRL) based algorithm named Modified Multi-agent Deep Deterministic Policy Gradient (M2DDPG), which is learned centrally and executed discretely. The simulation has demonstrated the validity, efficacy, and superior performance of the proposed M2DDPG algorithm compared to the baseline algorithm.

Permanent magnet synchronous motor has the advantages of small size, low loss and high-power density. It has been widely concerned and used more and more in real life. This paper proposes a permanent magnet synchronous motor drive system based on wireless power transmission, introduces wireless power transmission technology into the motor drive system, and further broadens the application range of permanent magnet synchronous motor. The main contents of this article include the composition and principle of each part of the system, builds the system simulation, and gives the simulation results. The results show that the system is feasible.

With the deepening of the depth of open-pit mining, the country's open-pit mining gradually into the deep depression mining stage, the number of high and steep rock slopes increased, landslide accidents frequently occur. Landslide accident has become a major hidden danger in open-pit mine safety production. Microseismic monitoring technology has been widely recognized by the industry as an effective slope monitoring and early warning means covering the whole mine. In this paper, a wireless microseismic monitoring scheme is formed through wireless bridge technology and solar power supply technology, which is successfully implemented in an open-pit mine slope monitoring. The data processing and analysis results are uploaded to the cloud service platform in real time. The wireless microseismic monitoring system based on cloud service platform runs stably, successfully solving the problem of long-term stable operation of the open-pit slope monitoring system and ensuring the safety of mine production.

With the iterative upgrade of electronic products, high precision band gap reference source is becoming the core of its development. Temperature compensation technology is the most important to design a high precision band gap reference source. In recent years, new high order temperature compensation technologies have been proposed. For example, segmental compensation technology, curvature compensation technology, and subthreshold compensation technology are common compensation means. In this paper, a new compensation method is proposed which is to generate the compensating current through the self-opening characteristic of the depletion-mode transistors. It can obtain a band gap reference source with high order compensation. The temperature coefficient can be as low as 3.3ppm/°C in the range of -40°C to 85°C.

Terahertz (THz) communication is a key technology to 6G wireless communication systems. Due to the extremely short wavelength of THz wave, ultra-massive multiple-input multiple-output (UM-MIMO) systems can be realized. The high array gain generated by UM-MIMO systems can overcome the limitation of THz communication distance. In UM-MIMO systems architectures, dynamic array-of-subarrays (DAoSA) architecture can dynamically adjust the connections between radio frequency (RF) chains and subarrays according to the channel state. In this paper, a hybrid precoding scheme of DAoSA architecture with low complexity and good performance is proposed. Hybrid precoding of UM-MIMO systems is optimized with the spectral-energy efficiency rate (SEER) maximization as the objective function. To reduce the computation complexity of optimization problem, the intractable original problem is decoupled into two subproblems that can be solved easily. Then, element-by-element (EBE) algorithm is proposed to solve the DAoSA hybrid precoding subproblem. Furthermore, to determine the connections of the switch es, a logistic chaotic searching algorithm (LCS) is proposed, which can obtain the global optimal solution without traversing repeatedly all connections under a fixed search space. Simulation results showed that the proposed DAoSA hybrid precoding scheme can achieve higher SEER than the existing papers.

To get multi-band antenna with reduced system load, we propose tri-band antennas utilizing shared aperture array which works in S-band, X-band and Ku-band, where three kinds of series-corporate microstrip antennas are interlaced on the flexible substrate. With the orthogonal polarization, layout of antenna arrays is optimized to ensure high channel isolation of antenna arrays’ different ports. The simulated results show the designed antennas having gain of 10.8 dBi, 11.8 dBi and 17.6 dBi respectively for S-band, X-band and Ku-band incidence. The side-lobe level (SLL) for S-band is -23.5 dB with return loss bandwidth of 50 MHz, while Ku-band antenna is -18 dB with bandwidth of 600 MHz respectively. And the bandwidth of circular polarization axial ratio (AR) less than 3 dB is around 1.1% for X-band. The isolation is better than - 40 dB among different ports of all arrays. Furthermore, the designed antennas can be flexibly covered on curved cylinder surface, maximum radiation gain is 2.4dBi, which is better than that of flat when the bending radius is 600mm for S-band. The proposed flexible microstrip antenna can be applied for conformal antennas, which shows great potential for flexible electronic device.

The consensus protocol in heterogeneous opportunistic networks has a wide range of applications in handheld device networking, vehicle networking, and formation control. At present, most of the distributed consensus algorithms are ideally modeled and consider a single factor, But in real life, heterogeneous opportunistic networks need to consider more complex factors, such as node dynamics, delay and packet loss generated by node information transmission process, and deviation caused by the unbalanced nature of the network topology. Aiming at the heterogeneity of heterogeneous opportunistic networks, we propose a quantized caching mechanism, which solves the consensus problem of heterogeneous opportunistic networks .And the optimal number of cache blocks of each node is obtained, which makes the convergence speed reach the fastest .Our main contribution is to provide the basic theory of consensus protocols for heterogeneous opportunity networks, and prove the correctness of the theory through experiments.

With the advent of the information age, a large number of information has emerged, and people need to deal with more and more information. In this case, text summarization technology came into being. As a Text summarization model, Transformer has the problem of obtaining key text information and out of vocabulary, which is difficult to effectively solve and resulted in the accuracy of the generated summary not achieving the desired effect, an improved transformer model was proposed, The model constructs an Aggregation module between the encoder and decoder to make the generated abstract more appropriate to the original content ,and the pointer network is introduced at the decoding end of the model to complete the decoding of the text summary, The function of pointer network that allows the unlisted words in the vocabulary to be copied directly from the source text is used to generate a text summary, which improves the flexibility of summary generation, Using the News2016zh dataset to test the performance of the model. In the experimental results, the improved Transformer model performs better than the benchmark model on ROUGE evaluation indicators.

Task offloading strategy is one of the promising methods in mobile edge computing (MEC) to reschedule computing resources. Scheduling for computationally intensive tasks was mainly focused on and the dependency restriction between tasks was generally ignored. In this paper, an End-to-Edge collaborative resource allocation model is established. Firstly, in order to improve the unrealistic assumption raised in previous studies that tasks can be offloaded to MEC servers with arbitrary proportion and order, we establish the directed acyclic graphs to describe the dependency relationship within tasks. And then, system time cost including execution time, transmission delay and waiting latency are considered, and the task offloading problem is transformed into an optimization model to minimize time consumption. To manage the non-convex problem, an offloading strategy under dependency constraints based on deep Q-network (DQN) is proposed. Simulations prove that the proposed algorithm can obtain smaller time cost by comparing with other baseline algorithms.

An exponential curvature-compensated bandgap reference(BGR), which is designed in 0.18μm CMOS process, is proposed in this paper. The Darlington transistor structure is applied to reduce offset voltage of Amplifier, and a CTAT current circuit is used to generate the current with the higher order. To meets the needs of ADC, the noise of BGR is reduced by RC filter. Simulated in the 0.18μm DBH technology, the proposed BGR can achieve a total output noise of 9.54μv between 0.1 Hz and 1G with a filter capacitor of 5pF. A temperature coefficient of 2.1-ppm/°C is achieved by using exponential curvature-compensated at typical corner, and the temperature coefficient is less than 5-ppm/°C in all processes.

As the latest generation communication technology featuring high speed, low delay and massive access, 5G provides an efficient means of communication for the construction of new power systems. Based on 5G technology and combined with the characteristics of power grid, this paper studies terminal security encryption technology, remote monitoring and management technology, network probe monitoring technology. The Application of compressed transmission technology in differential protection of distribution network is realized. Build-in network probe to realize real-time monitoring of network quality has promoted the application of 5G technology in the power industry.

To ensure the safe operation of a superconducting accelerator system, real-time monitoring of the RF power source, transmission lines, and superconducting cavities is essential. Currently, the main method for monitoring the status of transmission lines in superconducting accelerator systems is through monitoring the standing wave ratio. However, it is difficult to effectively monitor faults during high reflection or full reflection operations, which can pose significant safety risks. To address this issue, this paper proposes an online monitoring and positioning technique for RF transmission line faults based on acoustic fingerprinting. By studying the spectral characteristics and transmission mechanism of high-power RF transmission line faults, the sound recognition and classification experiment achieved a recognition accuracy of 98.0%, demonstrating the feasibility of this method in identifying faults in RF transmission lines.

Railway signal is the technical assembly to ensure the safety of railway traffic, how to intuitively observe the operation of railway signal equipment is crucial, two-dimensional state of the computer interlocking system obviously can no longer meet the current needs of the times, digital twin as a new technology, can be observed from a dynamic multi-dimensional, multi-scale perspective of the physical world, cognitive, this paper will be based on the computer interlocking system, through the digital twin. This paper will establish a station digital twin software based on the computer chain system, based on the establishment of this system, it can monitor the railway signal in real time intuitively; this system can show the working process and working principle of the station signal equipment in detail, and achieve the effect of three station redisplay, so it also has a positive effect on the training of station staff and students studying railway signal system.

With the rapid development of artificial intelligence, the demand for Graphics Processing Unit (GPU) resources is also increasing rapidly. To improve the efficiency and utilization of GPU resources, virtualization technology has been widely used in the field of GPUs. This article reviews the evolution from GPU virtualization to resource pooling, including device simulation, GPU pass-through, hardware-assisted virtualization, GPU full virtualization, GPU remote-sharing, and GPU resource pools. The advantages and disadvantages of each stage and the challenges faced during the evolution process are also analyzed. This article discusses the difficulties and solutions of GPU resource pool construction, as well as the construction steps and framework, providing a reference for the research and application of GPU resource pooling.

The main service mode of communication protocol or communication business support software for upper application is API (Application Programming Interface). This paper focuses on whether the defective API is abused. Based on the method of extracting the feature set of defective communication API, and the defective communication APIs called in the APPs are analyzed and the communication security portrait of the APPs is builded. 20 typical open source communication software are selected to extract the feature set of defective APIs, and 1000 candidate APPs that may have security risks are screened to further evaluate the risks to obtain their communication security profiles. The experimental results show that the methods proposed in this paper can effectively detect the functional defects and privacy disclosure defects in communication APPs. This method not only provides a new technology for communication software security testing, but also provides a new idea for communication software security protection.

The certificateless cryptography solves the certificate management and key escrow problems in the Internet of Things. However, these schemes are no longer safe in the quantum environment. Lattice cryptography has become the main method to solve quantum security. At present, only a few lattice-based certificateless signature schemes have been proposed and the use of the trapdoor function limits the efficiency of the lattice cryptography, and the security of these schemes can only reach existential unforgeability. We construct a lattice-based certificateless signature scheme based on Lyubashevsky's signature scheme. The security of our scheme can reach strong unforgeability and the computational efficiency of our scheme is lower than the previous schemes.

The full coverage of electronic products promotes the prevalence of human-computer interaction systems, especially in sports and fitness. Excellent human-computer interaction technology can bring high-quality experiences to human beings. However, the previous research focuses on the tracking of the wearable device and the radio frequency signal to the body part. This article will launch a lightweight, portable yoga private education system that can be mainly in line with coaches. A smartphone can realize real-time tracking of user body motion, combining the evaluation model feedback and valuable guidance. We predict the composition of the limb occlusion by the algorithm and use the segmentation to normalize the accurate calibration of the bone node. Improve the accuracy and robustness of the system while helping users achieve freedom and practical yoga training. The experimental results show that the system can provide meaningful feedback to the user in real-time by accurately assessing the user’s yoga action, high accuracy.

Joint entity and relation extraction is an important task in natural language processing, whose purpose is to obtain all triples in text. However, the existing models seldom pay attention to the part-of-speech (pos) of each word and the dependency parsing (dp) in the sentence. To solve these problems. a joint extraction model with part-of-speech-aware attention and dependency parsing embedding is proposed, named PADPE. The proposed model obtains better word representation through pos-aware attention mechanism. In addition, the parts of speech and dependency characteristics are integrated respectively in entity classification and relation classification to improve the accuracy of the classifier. The experimental results demonstrate that our model can solve the overlapping triple problem more effectively and outperform other baselines on three public datasets.

In natural language processing, the research on extracting triple relations from unstructured texts to construct knowledge graphs has been widely concerned, however, few existing methods can solve the problem of overlapping triples and pay little attention to relations interaction with sentences. Besides, because entities have multiple forms, most methods rely on contextual features during entity recognition, and these models often misrepresent global relations. These all affect the accuracy of triplet extraction. To address the above issues, this paper proposes a joint extraction model with specific relation embeddings as well as incorporating global relations. It is called SRGF. Specifically, the model first converts the text into a vector representation that combines contextual features and global feature vectors through a pre-trained model and a graph convolutional network GCN, and then extracts topics. So entity extraction can be enhanced. An attention network guided by subject and relation information is employed to learn improved sentence representations for identifying the related objects. By doing so, it becomes possible to capture and model the interactions among entities, relations, and sentences, thus leading to better solutions for the problem of overlapping triples. Experimental results on public datasets NYT and WebNLG demonstrate that SRGF outperforms previous methods in triplet extraction.

This study proposes an environmental risk perception and early warning method for metrology laboratories that combines information acquisition, feature extraction, deep learning, and bias analysis. First, various sensors are used to collect operating data from the laboratory and extract its operating characteristics. The data is then processed using long and short-term memory neural network (LSTM) to predict the laboratory's operational state. Finally, the generalized extreme value theory is employed to establish the alarm threshold based on the normal running state, enabling early risk warning of the equipment. The experimental results demonstrate that the LSTM model is highly effective, achieving stable risk prediction accuracy above 98%, surpassing other neural network models.

Since fiber Bragg grating has good reliability, stability, anti-interference and intrinsic safety, which is very suitable for the application of coal mine gas environment, a fiber Bragg grating stress sensing monitoring system based on metal bolt in coal mine is developed. This paper focuses on the design and research of the core equipment of the system, grating stress sensing demodulation device. Through the study of strain measurement principle of FBG and the study of demodulation technology of tunable F-P filter, through the design of hardware circuit and software program of demodulation device, the special stress demodulation device of fiber Bragg grating bolt is developed. Finally, the accuracy and practicability of the demodulation device are verified by the actual experimental test of fiber Bragg grating stress sensor.

This study proposes a feature extraction module named Feature Selector to enhance the generalization performance of neural networks for state detection of High-Voltage Switchgear (HVS) in outdoor environments. This module comprises three autonomous feature extractors, each intended for extracting distinct types of features, and a switch employed for choosing appropriate features for subsequent propagation. Experimental results on testing sets in three different environments demonstrate that the module effectively improves the generalization performance of the YOLOv5-nano (You Only Look Once) network.

The mode and principle of VIENNA rectifier are analyzed, and the current and voltage double closed-loop control scheme is selected by analyzing the characteristics of the system and mathematical modeling. The feedforward decoupling control method is used to reduce the errors caused by strong coupling and nonlinearity in the inner current loop, which increases the rapidity and accuracy of the current loop. PI control is added to the outer loop voltage loop, which reduces the output voltage fluctuation caused by disturbance and improves the stability of the system. Finally, MATLAB simulation verifies that the system can be fast and stable, which reflects the accuracy and rationality of the current and voltage double closed-loop control.

In order to solve the problems of low robustness and high complexity of power allocation algorithm in cooperative communication, under the constraints of limited channel strength, number of nodes and total power, this paper presents a power allocation algorithm based on channel capacity in cooperative communication. Based on the traditional channel capacity model, the harmonic mean and average weight of channel strength are introduced to simplify the solution process of the objective function in order to reduce the complexity of the algorithm. The crossover and mutation rules in the genetic algorithm are introduced to avoid local optima, and intelligent velocity constraints of particles are established to reduce the algorithm complexity. The optimal function is established by calculating the channel strength weight coefficients. Simulation results show that the proposed algorithm has improved in channel capacity and convergence compared to the mainstream and original algorithms. The improved algorithm has better practicality and advancement in cooperative communication.

To study the erosion problem of the shale gas well field, the CFD model of the elbow of rotating valve group was established to study the effects of gas velocity and sand content on elbow erosion. The research shows that the pressure in the valve group changes obviously at the elbow, and the pressure in the inlet blind pipe section is relatively high. When it flows through the tee area, the pressure distribution is uniform and the pressure difference is not obvious. Under the condition of constant pressure velocity and sand size, the maximum erosion rate of elbow increases gradually as the sand content increases. It is recommended to use downhole throttling to reduce sand volume and velocity to avoid serious erosion.

In the conventional encryption method, the encryption effect is mainly improved by expanding the encryption space. But the encryption randomness is large. The encryption security is reduced. Therefore, the subject of secure encryption method of privacy set intersection based on a homomorphic encryption algorithm is designed. By using the homomorphic encryption algorithm, the encryption privacy set intersection protocol is designed to encrypt the random elements of privacy combined with communication. The privacy set intersection encryption extreme difference protocol is established to prevent communication data from being stolen and ensure the security encryption effect of privacy information. By the way of comparative experiments, it is verified that the encryption method has high security and can be applied to real life.

At present, the public data platform lacks high-security certification for its behavior and access, and cannot effectively respond to internal and external threats, which can easily lead to security issues such as network security and data leakage. To this end, this paper designs a data access security monitoring system based on zero trust mechanism. The zero trust system is designed to consist of five components: Trusted Access Console (TAC), Trusted Application Agent (TAP), Trusted API Agent (TIP), Intelligent Identity Analysis System (IDA) and Trusted Environment Awareness System (TESS) . Through quantitative analysis, the security risk perception value of the public data platform is obtained, so as to achieve the purpose of security risk perception of the public data platform. It is proved by experiments that the system has high recognition accuracy and can effectively improve the average access bandwidth of the network.

At present, there is still seldom research on how to use neural networks for the surface defect detection of nylon yarn packages. The original VGG has some shortcomings, if there are many network layers, it is difficult to train, and if there are few network layers, it is impossible to learn abundant features to meet the requirements of industrial production. In this paper, an improved VGG-based method was proposed for detecting surface defects on nylon yarn packages. A two-path net