This PDF file contains the front matter associated with SPIE Proceedings Volume 12167 including the Title Page, Copyright information, and Table of Contents.

Abstract: To prevent communication failure in presence of severe underground coal-mine accidents, a large number of research was conducted to improve the effectiveness and reliability of the underground wireless communication systems in literature. In this paper, we first propose to replace the conventional lumped communication system, with the distributed network system. We also demonstrate a software-defined adaptive networking scheme. This scheme enables the automatic re-configuration, when some links or nodes fail in the network. At the same time, the routing strategy based on field strength is provided and the simulation verification is carried out. The results show that the proposed strategy can avoid the inefficient information exchange in the network, and significantly prolong the network life cycle. In addition, we propose an accumulative reception method, coupled with a multi-node co-transmission technique, that significantly improve the transmission reliability. The method is derived mathematically and verified by numerical simulations. The results show that the signal-to-noise ratio at the receiving end is improved, leading to a more reliable communication. The transmission throughput can be maximized when downhole accidents occur. These technologies lay a solid foundation for a more intelligent, efficient and reliable next generation underground wireless communication system.

Because the size of online files (videos, audios) increases, and longer distance of signals is needed to be transmitted, people today demand a faster and better-quality communication technology in long distance. Satellite communication is a technology which is developed last century and it is still used in particular areas, such as navigation systems based on satellites in different countries, satellite phones, satellite TV in remote areas, etc. However, satellite communication was replaced in most area by terrestrial communication because of its drawbacks. This paper is about the main advantages and disadvantages of satellite communication, and a few solutions, and discussions about the future of satellite communication. Researches on satellite, satellite communication, and terrestrial communication are adopted in this paper. The conclusion is that satellite communication will play an important role in the future, and some of its drawbacks can be avoided.

Aiming at the problems of small separation range and low separation rate in the intelligent separation of coal and gangue, a threshold separation method based on dual-energy X-ray technology is proposed. Low-energy original image, use ratio method to obtain R-value image, integrate R-value image and low-energy image information to fit the R-L curve of coal and gangue, count characteristic values and draw sample distribution maps, calculate the sorting threshold of coal and gangue, and realize coal and the sorting of gangue. The experimental results show that the threshold sorting algorithm is more effective in improving large-size materials than the traditional R-value recognition algorithm and R-L curve fitting algorithm, and the recognition rate of 150-300mm materials is as high as 96.99% and 92.32%.

An all-fiber temperature sensor is proposed for operational characterization, which is based on a Dual-Mode Fiber Bragg Grating (DM-FBG). This DM-FBG embedded in a Multilayer-Core Dual-Mode Fiber (MC-DMF) could sense the surrounding temperature depending on the transmission dips. The MC-DMF, with only two modes available for propagation, would make it easier to fabricate an FBG with two transmission dips. These two transmission dips have varying sensing precision. The propagation features and the working principle are described at length. With the temperature range of 30–70°C, the sensitivity of the DM-FBG, ~9.32 pm/°C and ~9.41 pm/°C are experimentally achieved, respectively. DM-FBGs could be scattered along a single fiber easily, and then a quasi-distributed sensing network would be built by these multi-node ones. Thus, this sensor has great potential for the internet of things.

Due to the complex aging characteristics of lithium batteries, it is an unsolved problem to accurately predict the health of the battery. This problem has also largely affected the development of electrical energy storage. Battery aging involves chemical, physical and mechanical factors, and it is difficult to establish a unified standard for accurate prediction. This paper establishes a method for estimating the capacity of lithium batteries based on data-driven and Gaussian models. By analyzing the relationship model between battery power generation and charging, the effectiveness of the method proposed in this paper is determined through relevant data.

Compared with cloud computing, edge computing shows real-time, security and other advantages when dealing with the explosive growth of the Internet of Things. In the process of data processing and mining data value, edge intelligence combines the advantages of edge computing and artificial intelligence to become a research hotspot. Federated learning is one of the most popular machine learning frameworks for edge intelligence. It provides stronger privacy protection for the client through the aggregation of the server-side client-side model. However, the latter often do not participate in the federal learning system due to resource constraints and economic considerations. Therefore, incentive mechanisms are necessary to attract higher quality clients. To this end, we use the Stackelberg game model to model the server and the client, and establish an incentive-driven federated learning algorithm FEDID. Finally, we verified the effectiveness of the incentive mechanism through experiments.

Nowadays, intravenous infusion has become the most effective medical means in clinical medicine. However, when having an intravenous drip, patients need to be monitored by medical workers. Therefore, this paper proposes an unattended intelligent intravenous infusion system. This system uses STM32 as the main control chip and realizes the detection and display of liquid level, the setting and adjustment of dripping speed, and real-time monitoring through JXWHW-005 infrared sensor, 0.96-inch OLED display, 28BYJ-48 stepper motor, keys, BT04A Bluetooth module, and USB-TTL conversion interface. This system has such advantages as safety, high efficiency, and saving manpower and material resources. And it also provides convenience for medical workers to monitor and manage patients.

The precise segmentation of breast tumors is of great significance to the clinical diagnosis of breast cancer. Automatic breast ultrasound (ABUS) is a low-cost medical imaging method. However, one breast scan will produce hundreds of three-dimensional ABUS slices. Manual screening for abnormalities is not only time-consuming, but also heavily relies on the doctor's experience. In this study, a new network with Orthogonal Positioning Attention Deep Supervision Network (OPADSN) was proposed for breast cancer tumor segmentation. OPADSN adopts an orthogonal strategy, sending the input to the horizontal axis and the vertical axis to output features in two independent directions. Orthogonal strategy not only retains more location information, but also helps the model locate and identify the target of interest more accurately. In addition, this article introduces the Original Resolution Subnetwork Network (ORSNet) into the context module for the first time to form the Original Resolution Subnetwork Enhancement Context Attention (ORS-CA). ORSNet can solve the problem of edge blur caused by downsampling, thereby generating high-resolution spatial features. Repeated use of the ORS-CA module can establish the connection between the breast tumor area and the boundary clues, making the tumor contour easier to judge. Due to this iterative interaction mechanism between regions and borders, OPADSN can correct different regions in the prediction results. This article conducted an experiment on a dataset of 783 two-dimensional breast ultrasound images. The experimental results show that the OLADSN method achieves 93.1% segmentation accuracy in Mean Dice, which is a satisfactory result.

Configuration design is an important means of experimental measurement of modern electric energy. Its appearance connects each link in the process of electric energy experiment measurement as a whole, greatly saves the cost of management, maintenance, application and other procedures, and also greatly improves the work and management efficiency of electric energy experiment. The measurement accuracy plays a crucial role in electric energy experiment. Based on the configuration design technology of computer, this paper studies the application of this technology in the precision measurement of electric energy experiment, and further analyses its specific application.

Under the influence of China's sustainable development strategy, the design of urban road traffic engineering should adhere to the concept of environmental protection and develop in the direction of green and modernization. This paper proposes a slow traffic optimization at intersections based on green building materials and signal processing technology. The green building materials, such as colored asphalt concrete pavement and polyurethane elastic isolation column, are used for the channelization design of the traffic space for slow traffic at the intersection. Then the slow traffic signal optimization processing technology is used that is the slow traffic signal timing algorithm model to optimize the slow traffic crossing time. Finally, the evaluation and analysis are carried out through VISSIM traffic simulation. The results show that the method presented in this paper can effectively reduce the average delay of vehicles and the average travel time of slow traffic at intersections.

In order to ensure the good state of the spare parts of the satcom servo system on board, it is very necessary to check the spare parts regularly. In this paper, a kind of off-line detection device for spare parts of satellite communication servo system on board is designed. It has high integration degree, many test interfaces, can be connected with monitoring software, easy operation, and has the function of fast and accurate test of spare parts, which is of great significance to the rapid detection of spare parts. After testing, the system can perform rapid detection of 14 spare parts in 8 categories at the same time.

Combined with the current situation of the country, by analysing the national policies on recycling and investigating the existing waste classification methods, a waste recycling method based on integrity and mobile communication technology is proposed. At the same time, it is put into practical application, which realizes the scientific classification of waste, improves the delivery efficiency, and ensures that waste can be classified scientifically and priced accurately, so as to realize human the benign and scientific sustainable development model of resources and environment.

Active defense is currently a key technology to reverse the asymmetry of offensive and defense in cyberspace. Honeypots, as one of the active defense technologies, are deployed in the internal network to attract attackers’ attacks, consume the attackers’ energy, and discover the attackers’ attack methods and attack intent. But at the same time, if there are security loopholes in the honeypot itself, the honeypot becomes the entrance for the attacker to attack the intranet, which will harm the security of the intranet. In this paper, aiming at the virtual machine escape scenario in the honeypot system, based on the mimicry defense idea, a dual mimicry mechanism and the honeypot architecture under this mechanism are proposed. This mechanism uses the heterogeneity of the underlying virtualization platform to resist the escaping vulnerabilities of the virtualization platform, and achieves level heterogeneity through honeypots to attract attackers, which is conducive to the complete collection of attackers' attack behaviors. Finally, the security test and performance test were carried out through the web implementation of the mimic honeypot.

Automatic three-dimensional breast ultrasound (ABUS) is an innovative method for assisted screening of veins in the lower extremities. However, reviewing the thousands of images produced by ABUS is not only time consuming but also easy to miss the venous vessels area with weak feature. In order to analyze the application value of deep convolution level set in ABUS image analysis of varicose vein of lower extremity, an automatic detection algorithm based on the deep convolution level set method is proposed in this paper. First, the experimental data is obtained through ABUS imaging of venous blood vessels. Then, deep convolutional neural network and level set method are combined for automatically realize the position detection and contour segmentation of veins. Finally, the three-dimensional reconstruction of venous vessels is performed according to the ABUS imaging physical parameters. Experimental results are shown that the combination of deep convolution neural network and level set method can obtain a clear and complete three-dimensional venous vascular network of the lower extremities, which has important clinical significance to assist doctors in formulating personalized treatment plans.

China's development of electric vehicles (EV) and distributed generations (DG) are conducive to getting rid of dependence on fossil fuels and realizing energy transition. However, the time scale between electric automobile charging load and distributed generator supply is difficult to keep consistent. Which will reduce the power supply reliability of the distribution network. This article presents innovation method to improve distribution network reliability use the interaction between electric automobiles and distributed generators. First, by analyzing the output characteristics of distributed generator, the intermittent power generation model of distributed generator is established; secondly, the charging habits of electric automobile owners are studied. The stochastic charging model of electric automobile is constructed; then based on the coupling relationship between electric automobile charging and distributed generator the establishment The power supply and demand balance model of electric automobile and distributed generator is developed; Finally, the IEEE-RBTS Bus-6 system is used for simulation, and the sequential Monte Carlo evaluation method is used to calculate the reliability indexes of the distribution network with different interactive response rates. According to the simulation results, the interaction between electric vehicle and distributed generator can effectively improve the reliability of distribution network.

In order to avoid reading errors due to artificial factors and improve work efficiency and effectiveness, taking the application of "machine vision technology" as an example, we design a smart converter station meter reading recognition system with perfect functions and strong practicability. First, the overall architecture design of the system is analyzed. Secondly, starting from the three aspects of image preprocessing, feature vector extraction, and reading recognition, the design of the digital meter reading recognition module is completed. Finally, starting from the server-side software system framework, system function module design, data structure and communication design, etc., the design of the server-side data back-end processing software system is completed. The results show that under the application background of machine vision technology, the meter-reading recognition system of intelligent converter station is operating normally, reliably and stably, and the realization of each functional module meets the relevant design requirements and meets the actual application requirements. It is hoped that this research will provide effective reference and reference for relevant practitioners.

The rapid improvement of neural network technology is now increasingly used in 3D modeling of furniture making it possible to quickly obtain 3D models from a single image. However, most of the current work is based on experiments on single furniture without background, and for complex interior backgrounds, the efficiency of reconstruction is greatly reduced. In this paper, we extract a system to extract a single furniture picture from a complex background and perform 3D reconstruction through the picture. Neural network-based instance segmentation and occupancy network techniques are used, and a suitable loss function is designed to evaluate the system. Through experiments, our method is significantly improved compared to other works and can effectively perform 3D reconstruction of furniture from complex interior scenes.

Taking the transportation system into consideration is a necessity in promoting the construction of safe and resilient cities. Select several types of indicators such as satisfaction with traffic conditions, public understanding, cognitive factors, countermeasure analysis, and confidence index, and carry out data analysis to understand the current situation of traffic congestion in Hangzhou based on specific research indicators. It is proposed that by constructing an intelligent transportation system, promoting the intelligentization of buses and new energy, connecting passengers, vehicles and big data. Establish an intelligent, digital and artificial intelligence-led traffic regulation network. Provide a theoretical basis for Hangzhou to build a resilient city.

Aiming at the problem of "difficult to put books on shelves" in the current library management, an intelligent book robot and its method of handling, sorting and loading are developed. It can autonomously locate, move, and avoid obstacles, and plan the path after obtaining the position of the book, and then pick up and transport the books, and place them on the bookshelf position. A new shelf device is created to grab stably most of the books and push them into the bookshelf. Its two-stage flexible expansion plate is developed to effectively expand the gap between books and have a high degree of tolerance. Mecanum wheels are used in the motion module, and the gray-scale sensor array are installed under the chassis in order to control effectively the posture of the car body. Book position locating is implemented by the visual recognition module with simultaneously recognizing multiple QR codes on the book spine. The results show that the rates of recognition and the completion of successful handling and loading are quite high.

The main source of noise in the substation is the power transformer. The radiated sound intensity of transformer noise is an important factor affecting boundary noise of substation. It is also an important parameter of equipment factory control. In this paper, the noise intensity of variable voltage under no-load voltage and load current is tested. The transformer noise testing datum is designed. The factory noise intensity characteristics of 220kV power transformer are obtained. The project provides accurate test data for transformer factory noise evaluation.

Environmental noise in cockpit can have great influence on the pilots of a helicopter. The pilots can be both mentally and physically influenced, which can weaken the competence of recognition and information processing and cause air accidents. In this paper, to further investigate the impact of environmental noise on cockpit pilot’s performance against risk information during a flight, experiment is designed and implemented in different noise levels and different task difficulty. During experiment the response time and accuracy of subjects are measured and recorded, the noise annoyance of subjects is evaluated by a subjective questionnaire. The experimental data show that with the rising of environmental noise level, the response time increases and the accuracy declines in all task difficulty condition. Also, the noise annoyance of subjects grows sharply in higher noise level. Based on the experimental data and result, it can be concluded that the environmental noise has a remarkable negative impact on the performance of helicopter pilot when performing a task and/or facing risk information.

In order to further improve the refinement of power grid planning, improve the quality of enterprise operation and economic benefits, a power grid planning method that considers the uncertainty of source and load is proposed in view of the uncertainty of distributed photovoltaic output and load forecasting. First of all, in view of the uncertain characteristics of distributed photovoltaic output, the concept of "photovoltaic output shielding factor" is innovatively proposed, and a probability distribution model of its daily and seasonal characteristics is established to analyze the spatial correlation of photovoltaic output. Then, on the basis of considering the investment cost of new lines, the cost of network loss, and the loss of light abandonment, a grid planning model considering new energy access is established. Finally, an empirical analysis is carried out through a typical power grid, and the results show that the model and algorithm proposed in this paper are effective and feasible.

Strengthening the lean management of the cost of operation and maintenance of distribution network equipment is one of the important ways to ensure the sustainable operation and development of power grid enterprises. Therefore, this article combines the relevant theories of grid management of the distribution network and analyzes the main factors affecting the operation and maintenance cost of different grid equipment based on the fishbone diagram method; and then combines the analytic hierarchy process, the entropy method and the grid comprehensive state evaluation and constructed an optimal allocation model of operation and maintenance inputs; finally, empirical analysis was carried out with 11 grids in a certain place as the research object, and the effectiveness of the model was verified.

At present, thanks to the development of related technologies such as computers and communications, the existing ubiquitous power Internet of Things related technologies have made considerable progress, such as widely supporting distributed energy access, energy optimization configuration, energy interconnection and sharing, and energy supply and demand balance. At the same time, the ubiquitous power Internet of Things is increasingly carrying differentiated energy services internally and externally. It is urgent to use virtualization technology to realize the concept of virtual operators, and provide and deploy services in the form of virtual service resource leasing to meet business needs. The operation requirements of ubiquitous power Internet of Things are not only isolated from each other but also integrated with each other. Existing energy operation management and control technology research is mostly carried out for decentralized networks. However, due to the heterogeneous, decentralized, and obvious differentiation of distributed energy, it is difficult to effectively organize and control the business, and it has not yet obtained better applications. Moreover, the ubiquitous power Internet of Things is a fusion model formed by the "cloud, management, edge, and end" ICT business function chain. Currently, there is still a lack of research on the operation and control technology for its unique resource chain mode.

At present, some achievements have been made in the field of Bluetooth, but the technical level needs to be improved. Relevant companies summarize the Bluetooth technology in the wireless communication field involved in their products in order to overcome the bottleneck problem of Bluetooth all over the world. The purpose of this paper is to analyze the current situation of Bluetooth technology, such as high power consumption, unstable technology and low security index, so as to broaden its application scope and field, so that it is no longer limited to household appliances, and can be applied to automobiles, information points, aviation, consumer electronics and military in the future. The data of this paper comes from the report of today's Bluetooth market. The conclusion is that Bluetooth will shine in the field of the Internet of things and wearable devices in the future, which is no longer limited to our previous research field.

Nowadays, multiple-input and multiple-output (MIMO) technology has been widely applied to the field of wireless communication and the generations of cell phone systems, especially in the systems of LTE-4G and 5G. It’s one of the several key technologies in this field and making a significant contribution to the improvement of innovation of generations of wireless communication. Firstly in this paper, it will include the basic analysis of some simple technology of MIMO technology including the basic cognition of MIMO. In quick succession, the principles of several related technologies will be analyzed followed by, including the carrier aggregation, the description of the whole process of the information processing in MIMO with analysis in details of each step like the formation of codons, the different kinds of layer mapping in order to correspond to precoding, the different antenna ports which represent different data streams, the classification of implementation effects in MIMO which include space diversity, spatial multiplexing and beam forming, the different modes of information transmission, and some major values about how to choose the mode of transmission.

Through analyzing the data characteristics of the typical business of the substation, an in-depth analysis of the need for safety isolation in the substation is carried out. After studying the change of substation demand and safety requirements, the model of open platform and specialized APP of monitoring system was proposed, then a three-in-one integrated management basic platform was designed to realize the integration of scientific modeling, data collection and data management of the whole station information. The platform provides standardized service access interfaces for resource management, data management, public services and application management, enabling flexible expansion of various applications. It effectively utilizes its own technical advantages in its professional field, makes it more convenient for functional applications to be integrated into the monitoring system and makes it more promotes the intelligent development of substations.

There are many devices in the smart distribution network, and the current differential protection methods are stronger for fiber optic or Ethernet communication, which cannot solve the problem of low protection efficiency caused by the protection data transmission delay. To solve the problems analyzed above, the adaptive differential protection method for smart distribution network under the framework of 5G communication is studied. After establishing the 5G communication framework for differential protection data synchronization in smart distribution networks, the protection data synchronization is realized by using the self-synchronization algorithm. The adaptive differential protection is achieved by adding braking coefficients to the conventional differential protection scheme and introducing new adequate protection criteria. In the simulation experiment, the protection efficiency of the protection method using 5G is improved by nearly 45.4%, and the effective protection rate is significantly increased.

With the popularization of self-service water dispensers on campuses, the way to identify users has become extremely important. In order to improve the identification method, fingerprint recognition is adopted in this design, which transmits data through ESP8266 module with STM32F103ZET6 as the processing core, stores information in the MySQL database of the upper computer, and uses Hall sensor YF-S201to collect water flow to help calculate the charges. WiFi transmission, water bill calculation, fingerprint recognition, RF card identification, and other functions of water dispensers are realized. This design not only facilitates the administrator’s management of users, but also ensures the security of user information.

This paper designs a transmission method of uncontrolled isolation RS485 communication. It is realized by using the differential transmission characteristics of RS485 bus and the transceiver characteristics of conversion chip. This method does not need CPU program control. It is convenient debugging and low cost. The method adopts optocoupler to isolate TTL signal. It meets the requirements of coal mine equipment for intrinsic safety and non intrinsic safety isolation of RS485 communication bus.

In order to obtain the transmission gain and beam pointing required by the system, the waveform signals in all channels of the radar transmitting system must ensure strict timing synchronization. For the multi-channel digital receiving system, the main factors affecting the synchronization are the synchronization of working clock, the synchronization of system reference signal and the synchronization of data transmission. This paper introduces the design of synchronization system for multi digital sub-array. Time-synchronization module combined with passive distribution module and phase stable cable is used to realize the generation, amplifier and distribution of LO signals, sampling CLK signal. At the same time, the sync trigger signals are generated, delayed and distributed. The synchronization system for multi digital sub-array has strong practicability and engineering reference significance.

Convolutional neural network models have become one of the most commonly used methods for analyzing medical images. Among them, the codec structure has brought important breakthrough results for medical image segmentation. However, the current medical image segmentation method based on the codec network architecture still has many problems. The corresponding feature map of the codec network in the skip connection structure has a large semantic ambiguity, which may increase the difficulty of learning the network and reduce the segmentation performance. The codec network architecture cannot make full use of the relationship between objects in the global view, and also ignores the global context information of different scales. In this article, we add attention gate mechanism (AGs) to the jump connection structure, and introduce attention mechanism and multi-scale mechanism to solve the above problems. Our model obtains better segmentation performance while introducing fewer parameters.

In order to realize the better application of 3D reconstruction technology in architecture, geological detective, meteorology and other fields, an improved surf algorithm is introduced to realize the reconstruction. By obtaining effective 3D points, the linear method and reweighted midpoint method are combined into triangulation. Finally, the point cloud matching and texture reconstruction algorithm are combined to realize multi view 3D reconstruction. This method only takes RGB image as input, does not need precise instrument, and the reconstruction time is much shorter. Experimental results show that compared with the traditional single view based 3D reconstruction, this algorithm has great advantages in both the number of point clouds and triangular mesh patches, and the 3D effect is clearer.

With continuous perfection of the BeiDou Navigation Satellite System (BDS), research on BeiDou positioning system is more and more in-depth. In order to study the civil application of the system, in this paper we model the positioning accuracy of the second generation of BDS (BD2) based on the PVT (Positioning Velocity and Timing) solution of Least Squares, and the simulation results show that, the PVT solution can improve the positioning accuracy in fewer stars, meeting the application requirements of civil vehicle navigation)

As one of the basic components in industrial systems, the safety and reliability of centrifugal pumps are directly related to production efficiency. This paper presents a fault diagnosis method of centrifugal pump based on EAS and stacked capsule autoencoder. First, use Electrical Signature Analysis (ESA) to select electrical signals as fault parameters for the fault data of the centrifugal pump; secondly, normalize the motor torque data of the six faults to the interval [0-255] and convert it into grayscale Images are input into the stacked capsule autoencoder network for fault diagnosis training, and selfattention-based pooling is used to reduce the number of capsules and increase the calculation speed. Train the Part Capsule Autoencoder (PCAE) to maximize the likelihood of the original image and the reconstructed image, and train the Object Capsule Autoencoder (OCAE) to maximize the likelihood of the original part and the mixed part, to obtain the optimal fault diagnosis model, and the classification accuracy of the optimized model is 96.57%. The method proposed in this paper solves the problems of complicated installation of fault signal sensors and poor generalization in fault diagnosis of centrifugal pump and improves the robustness and accuracy of fault diagnosis of centrifugal pump.

The flush temperature performance of smart toilets has a decisive impact on health and user experience. Aiming at the problems of complex testing process and simple data processing of existing smart toilet water temperature performance detector. An smart toilet water temperature detection system based on STM32 controller and Qt software platform is designed and implemented. This system adopts the Pt1000 precision platinum resistance sensor combined with constant voltage source circuit, signal amplification circuit, A/D conversion circuit, can realize accurate read the temperature signal, and through the look-up table method and the linear difference algorithm for temperature data for further fitting, finally will be sent to the measured temperature - time curve PC software features analysis and curve display. The experimental results show that the performance of the water temperature detection system is stable, the detection accuracy is not less than 0.5℃, and the detection efficiency is greatly improved compared with manual detection.

Aiming at the noise radiation and propagation problem in the external sound field of power reactor, a simplified reactor model was established using ANSYS workbench, and the simulation model was imported into the acoustic software LMS virtual.lab, and the finite element mesh was divided by the direct boundary element method, the vibration of the reactor surface is simulated and calculated, finally, the characteristics of the noise radiation and the sound field are analyzed, and the noise value around the reactor is predicted to show that the sound diffraction phenomenon is very strong, the sound pressure value changes rapidly, and the relevant results can provide a corresponding theoretical basis for reactor noise reduction.

A review of current research of sliding mode variable structure control is introduced, including the history of the development and basic definitions, and then discussed the main research directions and application field of the sliding mode variable structure control, which emphasizes on the application status of sliding mode variable structure control on spacecraft and servo system. Further more, the research tendencies in this field are discussed in this paper.

In order to reveal the characteristics of water gas two-phase seepage in soil pores and explore an accurate and efficient research method, this paper selects undisturbed soil as the research object, and establishes the three-dimensional model of real undisturbed soil samples by using computer tomography and image processing technology. Taking the simulation of water-gas two-phase seepage field as an example, the role of computer application technology in soil seepage research is discussed in detail. The results show that the application of computed tomography and image processing technology can better realize the dynamic visual process simulation of water-gas two-phase displacement in soil pores at meso-scale. This research method enriches the existing three-dimensional reconstruction simulation methods of soil mass, and provides a new way for meso-pore structure and fluid seepage research.

To avoid the hidden danger of some experiments and fill the lack of experimental teaching, a computer chemistry virtual simulation experiment system was built based on a unity3d platform and 3ds Max modeling technology. The results show that the overall satisfaction of students with the virtual experimental system is high, and more than 60% of the students are very satisfied with the use of the experimental system. The satisfaction of module design is the highest, and the satisfaction of interactive operation is relatively low. The dimension of students' interest in using the virtual experimental system is 3.946 ± 0.381, compared with the students under the traditional teaching method, Learning enthusiasm has been significantly improved. The real simulation effect of the virtual experiment can effectively improve students’ interest in chemistry experiment.

The application of computer technology in e-commerce not only drives the innovative development of computer technology itself, but also promotes the digital process of e-commerce, and promotes the innovation and development of computer technology in China to a certain extent. The most important part of computer technology is software technology. Computer software technology includes database technology, operating system technology, algorithm technology and data structure, information security technology, software programming technology, software testing technology and so on. Software programming technology in e-commerce is mainly to install software programs, system management and program writing. This paper mainly analyzes the specific application of algorithm, data structure and software engineering in e-commerce.

In the age of virtual currency's prosperity, the privacy of virtual currency cannot be ignored. By comparing the degree of privacy of Monero and Bitcoin and analyzing the technical background of Monero, it is found that the encryption method and privacy of Monero will become a major mainstream trend leading to the digital currency in the future. Through the discussion of the privacy, security, and non-traceability of Monero, we found that the security of Monero is obvious to all. The forecast of future market analysis compares the future development trend, market supply, and profitable rate of return of Monero and Bitcoin, confirming that Monero's market outlook is very impressive. Facing the emergence of digital currencies such as Zcash and Dash, Monero will still become the mainstream currency that will lead the development of digital currencies in the future.

How to automatically obtain cross-features with different weight values is a significant issue in the research of recommendation models. Traditional recommendation models cannot automatically learn the deep-level features of users and items to obtain cross-features. The mixed processing of dense numerical features and sparse categorical features will result in more information loss during dimensionality reduction. Cross features occupy the same weight in the recommendation process, which will lead to the non-prominence of critical features and reduce the accuracy of model recommendations. This paper proposes a personalized recommendation model (MSRN) for self-attention perceptron with automatic feature correlation. The model first processes the numerical features and category features in double towers to reduce the loss of feature information. Numerical cross-feature matrix and category cross-feature matrix use multilayer perceptrons to automatically mine the hidden knowledge and relationships between features. The model uses the Hadamard product to process it to obtain the cross feature matrix and uses the self-attention mechanism to assign different weights to the extracted cross-features. The experimental results on the public data set show that the recommended evaluation indicators of this model, MAE, and RMSE, are better than the current advanced recommendation models and have better accuracy and stability.

Information literacy (IL) is the basic literacy of undergraduate students. This study obtains 107 pieces of literature on the information literacy research of undergraduate students in the web of science core collection database, and uses CiteSpace software to analyze its annual publication numbers, journals, countries and institutions, keyword clustering, and topic changes. The results show that scholars have conducted many studies on undergraduates’ information literacy, and the annual number of publications has become an overall upward trend. The country with the highest contribution to this research field is the United States, and the institution is the Granada University of Spain. Information literacy competencies, information literacy courses, and undergraduate education have become the focus of this research field.

Pre-trained language model has a good performance in text summarization task, thus we present a neural text summarization based on a powerful pre-trained language model GPT-2. In this paper, we propose a Chinese text summarization model by extending into our downstream task to acquire relevant, contentful, and coherent summarization. By extensive experiments, our model achieves absolute improvements of 10.75% on ROUGE-1, 13.85% on ROUGE-2, and 9.73% on ROUGE-L on the LCSTS datasets. Compared with the state-of-the-art summarization model, e.g. BERTSUM based model, our model also achieves an improvement of 25.22% on ROUGE-1.

All work of library is carried out around the readers, library visits prediction is crucial to the allocation and optimization of library human, financial, and material resources. Based on the construction of GM (1,1) model and the Back Propagation Neural Network (BPNN) model, this study established a combination model of GM (1,1)-BPNN, and takes Jiangan Library of Sichuan University as the case study, and fits the 36-month data from January 2017 to December 2019 for model verification. Results show that, in terms of library visit prediction, GM (1,1)-BPNN has smaller errors, higher prediction accuracy, and better stability than GM (1,1) or BPNN.

The calibration of Automated Test Equipment(ATE) is an important part of microelectronics measurement. To achieve full calibration of high-speed ATE, each high-speed digital channel should be traced to the calibration equipment. This paper provides a design of high-speed signal switching system, which can effectively reduce or even eliminate the impact of path impedance discontinuity caused by the change of switching impedance, protect the safety of equipment and the stability of the signal to be measured, and effectively reduce the wave reflection effect caused by the change of impedance at the moment of channel switching.

As an important means of laboratory digital transformation and intelligent upgrading, digital twin technology has been widely used in laboratory upgrading and transformation. Digital twin technology can be used to digitally transform laboratory equipment in automated production lines, construct a digital twin virtual simulation experiment teaching system, and complete the transformation and upgrading of laboratory equipment. The results show that the design and transformation of the virtual simulation control system of a typical intelligent manufacturing production line meets the learning requirements of automation, manufacturing, and mechanical students. The construction of a practical teaching laboratory combining virtual and reality reduces laboratory construction funds and enriches the content which the students will learn.

The area along The Belt and Road is located in the Mediterranean Sea-the Himalaya Seismic Belt and the Pacific Rim Seismic Belt, which is one of the regions with the most frequent earthquake disasters and the most severe losses in the world. The establishment of The Belt and Road earthquake international cooperation fund can help maintain regional society peace, stability and prosperity. Based on The Belt and Road earthquake data of EM-DAT，this paper uses the neural network model VaR method to calculate the fund size of The Belt and Road earthquake international cooperation fund under different risk tolerances. According to the research results, when the risk tolerance is 90%, 95%, 99%, and 99.5%, the initial fund sizes are predicted to be 250.8 million U.S. dollars, 444.8 million U.S. dollars, 633.2 million U.S. dollars, and 910.8 million U.S. dollars, respectively.

In order to study the influence of cement-soil mixing pile arrangement in soft-soil foundation treatment on frame seawall, this paper took Wenzhou Lingkun seawall project as an example to analyze the effective cement-soil mixing pile length and spacing based on Midas GTS finite element analysis. The results turn out that the optimal cement-soil mixing pile length was 15~20m and the optimal pile spacing was 2.5 to 3 times the pile diameter on soft-soil foundation treatment.

Energy coupling multiple distributed clean energy, the Internet is to solve the current shortage of urban energy consumption and environmental pollution problems, one of the important means to research the coordinated development of energy and wisdom Internet city internal mechanism and application of map for supporting urban energy transformation, promote energy revolution, carbon neutral strategic target has important practical significance. Therefore, this paper firstly studies the relationship between the collaborative development of energy Internet and smart city from three levels of smart governance, smart industry and smart consumption. On this basis, the application map of the collaborative development of energy Internet and smart city is studied by combining the three levels of energy Internet equipment layer, platform layer and application layer. Finally, combined with the current carbon neutrality strategic goal and the new background and requirements of building a new power system, the future direction of collaborative development of energy Internet and smart city is proposed, in order to provide some reference for the planning, design and suggestion of smart city under the current background.

Using programmable NIC to offload network processing tasks such as checksum computation, virtual switching has become a common case in data centers. Such offloading mechanisms is able to help saving CPU cycles on servers, thus allowing data centers providing more computing resources to users for commercial benefits. However, recent research indicates that even with the offloading features above, data center servers still suffer from a great portion of CPU time being devoted to computing tasks that are unrelated to tenants. Such tasks include but not limited to data encryption and decryption, compression and decompression, etc. These tasks are common in modern data centers and cost a high amount of computational resources. Thus, in this paper, we propose a framework to offload those tasks onto programmable NIC, which will avoid high CPU usage on data center servers. On the one hand, the proposed framework is able to support multiple computing tasks running on the same programmable NIC at the same time. On the other hand, with a novel indexing mechanism, users are able to determine the order of different tasks at compile-time. Last but not least, the feasibility and advantages have been validated via our early stage experiments.

In abdominal hernia surgery, accurate detection of the Lightweight (LW) mesh has critical clinical significance for the diagnosis and treatment of mesh-related complications. Reviewing the large number of slices produced by Automated Breast Ultrasound (ABUS), however, is not only time-consuming and laborious but also easy to miss or misdiagnose the micro-structured LW mesh near the fascia tissue. Therefore, in this paper, an automatic and accurate computer-aided detection system based on deep convolutional neural networks and the level set method is proposed to improve this review. Firstly, the ABUS image is pre-processed using an intelligent speckle reducing anisotropic diffusion (ISRAD) to enhance the edge details of the LW mesh while reducing speckle noise. Then, combine the ROI prior information output by the deep convolutional neural networks and the level set method to outline the contour of the LW mesh. Finally, 3D reconstruction, and analysis of the LW mesh changes over time. The LW mesh with different imaging time, sizes, degrees of aggregation (DOA), and imaging depths are used to test the proposed method, experimental results show that the proposed method has a satisfactory application for detecting and analyzing the LW mesh in ABUS images.

To address the topological instability in the amorphous flat air-ground wireless self-organizing network caused by the high-speed movement of the network nodes, this paper adopts the 802.15.4 protocol as the basis of research, and uses dynamic planning in optimal control theory to backpropagate and optimize the directional steering angle of the mobile trajectories of the air and ground self-organizing nodes, so as to maximize the elimination of the wireless signal coverage blind area and enable as many self-organizing nodes as possible to form a stable, effective and reliable self-organizing network topology. This can maximize the elimination of wireless signal coverage blind areas and enable as many selfassembling nodes as possible to form a stable, effective and reliable self-assembling network topology. The simulation results show that by using the dynamic planning strategy, the optimized steering angle of the moving trajectory direction corresponding to the self-grouping nodes at different moving speeds can be obtained, and the sum of squares of the steering angles of the instantaneous moving trajectory direction of the optimized nodes can be obtained. The experimental results show that three fixed ground self-grouping nodes, one vehicle self-grouping node and one air self-grouping node are used to test the topology of high-speed mobile self-grouping network, and the five nodes can finally form a stable and effective self-grouping network topology.

In recent years, as the blockchain and the cryptocurrency built on it have become popular, a large number of decentralized financial applications have been built on the Ethereum network. This makes the security issues on Ethereum attract more and more researchers' attention. Phishing scams on Ethereum have caused people to suffer huge economic losses. Recently, with the popularity of graph convolutional neural networks (GCN), many models based on GCN for node classification have emerged. However, these current GCN models are difficult to cope with the challenges caused by the lack of side information and labels of nodes in the Ethereum network. In this paper, we propose a semisupervised graph convolutional neural network model based on important neighbors for the identification of phishing scam nodes on Ethereum. In our work, we design the pretext task for the node embedding module so that our model can learn the appropriate node embedding by using a large amount of unlabeled node data. Subsequent experiments show that our proposed model is better than all other baselines, which proves the effectiveness of our model.

The metamorphic test is a method to alleviate the unexpected value of the Oracle problem. The key point is the identification of the metamorphic relations. The identification of the metamorphic relations of scientific calculation programs is also a complex problem, and the likely of the metamorphic relation of the program can provide enlightening information for the identification of the metamorphic relations. The likely metamorphic relation can be regarded as the implicit expression of the input pattern and output pattern. This paper proposes an output pattern recognition technology based on the likely metamorphic relations of GEP. The technology is mainly aimed at the core neutron diffusion calculation program. The input pattern of the program, and then generate input data and run the program. Finally, in the corresponding output data results, through GEP data mining technology, the output pattern expressed in a variety of functional forms is obtained, which is further compared with analytical solutions and verified to be reliable likely metamorphic relations.

Existing modifications to the original (2D) Playfair cipher have made significant progress on improving its security, but little attention was given to optimize or maintain the original time and space efficiency. To close the gap, this paper proposes Hybrid Playfair, a modified Playfair cipher using a three-dimensional key and 2-character message pairs for encryption and decryption. With a schema made of five different substitution policies covering seven distinct cases, as well as an implementation guideline with sorted entry map and cross-layered key storage, Hybrid Playfair can offer not only better protection against frequency analysis attacks than 2D Playfair but also lower space and time usage than both 2D and 3D Playfair cipher. Hybrid Playfair can be used for applications on less powerful platforms with high throughput demands, such as live streaming and audio filtering devices, with its high performance-cost ratio and unimpacted security.

The tourism information management system design mainly adopts the object-oriented data model, where the technical key points of the system design can be found. Through ORDBS, the maintenance and modification of the information management system are easier. With the establishment of the tourism information management data model and transforming the model into a relational database, the tourism information management system can be better built.

Aiming at the low accuracy of elderly speech emotion recognition, we propose an emotion recognition method that integrates the elderly's speech features and embeds the attention mechanism in this paper. The method firstly extracts the speech features of the elderly and fuse them. Then the fusion features are used as the bidirectional long and short-term memory network (BLSTM) input to learn the deep emotional features of each frame of speech. The attention mechanism uses to calculate the weight of the emotional classification of each frame feature. Finally, the features of each frame multiplied by their respective weight coefficient are used as the fully connected layer input to complete the recognition of speech emotion. The experimental results on the elderly speech emotion database (EESDB) show that compared with the traditional BLSTM, this method can effectively improve the accuracy of elderly speech emotion recognition.

The anti-misoperation rules of substation are written by professionals, which are generally described in the form of text and contain complex semantics. It is difficult for non-professionals to understand, which is not conducive to reading, viewing and debugging of anti-misoperation locking logic. Aiming at the above problems, this paper proposes a visualization method of anti-misoperation rules based on real-time verification and simula-tion verification. Firstly, the anti-misoperation rules of substation primary equipment is analyzed, which is transformed from text form to binary form. Secondly, realize visual display and real-time verification through the combination of primary equipment wiring diagram and real-time database. Finally, the simulation database is formed by tem-porarily copying the real-time database to realize the visualization and simulation verification of the whole station anti-misoperation rules. Based on this method, the semantics and symbolic description of complex logical rela-tions can be shielded, and the logical relations can be visually displayed in a graphical way, so as to improve the audit efficiency of substation staff locking logic.

Aiming at the problem that the storage resources are challenging to use due to the fragmentation of storage resources of the Internet of Things terminal equipment group, this manuscript proposes a storage resource management strategy of minimum difference fragment storage. This manuscript builds a distributed storage architecture for terminal equipment groups among IoT terminal devices in the same local area network to realize file access and sharing among various terminal devices. When storing files, the minimum difference fragment storage method is adopted. Large files are stored in the device in a fragmented storage method; small files are stored in the device that the difference between the available storage resource size of the device and the file size is non-negative and the smallest. The simulation experiment results show that the strategy proposed in this manuscript can effectively improve the utilization of terminal equipment storage resources.

Face recognition has been widely used in information security, access control, financial payment, criminal investigation, and other aspects due to its stability, convenience, ease to forge, and other advantages. The Gaussian blur, a fuzzy algorithm, uses the Gaussian distribution to increase the efficiency of face detection. However, the data distribution of the Gaussian blur model is not effective enough for face recognition. In this paper, we improve the one-dimensional Gaussian kernel function to the two-dimensional Gaussian kernel function and use the Gaussian distribution to improve the accuracy of the recognition system. We improve the Model and adopt the Gaussian calculation method. According to the one-dimensional Gaussian function and two-dimensional Gaussian function to carry on the analysis and operation, the conclusion is drawn that the two-dimensional Gaussian function can enhance the efficiency of face recognition when applied to the designed face recognition system. To verify the effectiveness of our method, we compare the proposed method with Gaussian blur on three different datasets. The experimental results show that our method significantly outperforms Gaussian blur. Our analyses illustrate that the recognition degree and range of the improved model are more comprehensive and broader, decreasing the time used for recognition.

Space-Air-Ground Integrated network (SAGIN) that integrates satellite system, air network and ground communication network has attracted extensive research interests in recent years, for its high value to practical services and its wide applications in communication. Nevertheless, it also faces many unprecedented challenges due to its heterogeneity, selforganization, time variability and other characteristics. In order to solve the problems of unstable inter-domain neighbor relationship, frequent routing update and slow routing convergence in space-air-ground integrated network, a link fault recovery method of space-air-ground integrated network based on time sequence link weight graph is proposed in this paper. This method is based on the infrastructure of software-defined network, and introduces time-varying characteristics, designs a dynamic model of link weight change, and proposes a link fault recovery method. The simulation results show that by considering link resources and node resources, compared with the link detection and recovery scheme of softwaredefined satellite network, this method can effectively solve the problem that it is unable to build an effective recovery path because of the unstable inter domain relationship in the space-air-ground integrated network, and can effectively find a recovery path with the lower path cost, low end-to-end delay and high reliability.

Affected by uncertain factors such as equipment automation transformation and function upgrading, mechanical products may face varying degrees of upgrade risk. In order to solve the problems about high complexity of product upgrading, the optimization method for module partition considering upgrade risk is proposed. Firstly, the basic principles for module partition are analyzed to build fuzzy correlation matrix. With the goal of optimizing module partition quality and minimizing the upgrade complexity, the improved artificial bee colony algorithm is proposed to solve the model. Chaotic dynamic weight factor is introduced to enhance convergence speed of the algorithm. The set of non-dominated solutions for module partition is obtained. Finally, the module partition of an elevator is given to verify the effectiveness of this work.

In order to manage the aerospace product manufacturing, a quantified evaluation of product manufacturing readiness based on BP neural network is proposed. In view of the unique problems of Chinese aerospace product manufacturing, the risk factors of aerospace product manufacturing are analyzed, and each manufacturing factor is decomposed hierarchically to establish a three-level indicator hierarchy for the aerospace product manufacturing readiness evaluation. According to the evaluation indicator of manufacturing readiness, the qualitative indicators and the quantitative indicators are quantified according to the demand satisfaction and fuzzy mathematics membership function. Based on the BP neural network, a quantitative evaluation of Chinese aerospace product manufacturing readiness is modeled. The comprehensive scores of manufacturing readiness is calculated by BP neural network according to the indicator evaluation scores, then the manufacturing readiness level is evaluated quantitatively and objectively. In order to optimize the evaluation model for the aerospace product manufacturing readiness, trainrp and trainlm are selected as the training function respectively for training. The error analysis experiments show that the average relative error of the manufacturing readiness evaluation model using trainlm as the training function is small, which can provide a scientific method for the objective evaluation of the aerospace product manufacturing readiness.

The removal of speckle noise in synthetic aperture radar (SAR) images is important for the subsequent processing and analysis of SAR images. In order to suppress the speckle noise and improve the equivalent number of looks (ENL) in SAR images, a de-speckle noise model based on modified wavelet decomposition and partial differential equation (PDE) is proposed. In this study, the de-speckle noise model based on modified wavelet decomposition and PDE is compared with wavelet transform (WT) model, integer order total variation (TV) model, fractional order total variation (FTV) model, and tight frame (TF) model. The experimental results show that the improved total variance model has better de-noising effect on real SAR images and retains edge details, which is of practical value.

The selection of the benchmark network in self-supervised monocular depth-based estimation can often only be made using previous networks to select the best performers among them. When there is a change in resources and want to scale the network, it is difficult to find a suitable way to adjust the network quickly if the selected network itself does not give the same series of networks of different sizes. In this paper, we investigate whether the network generated by the Neural Architecture Search method based on search parameter scaling has good robustness in monocular depth estimation based self-supervised, as which the pose estimation network as well as the depth estimation network can have a better improvement in the accuracy of depth estimation. The final experiments show that the generated series perform well on the KITTI dataset, with the best performing EfficientNet-B3 outperforming all previous self-supervised networks.

Indoor and outdoor images captured by the camera are affected by atmospheric dust, haze, sand and other factors, resulting in graying of the images. Although existing image dehazing algorithms can achieve removal of haze, they suffer from incomplete dehazing and color distortion to some extent. To address the above problems, this paper proposes a singleimage dehazing network with multi-scale feature extraction. The network algorithm is based on recurrent generative adversarial network, and a feature extraction network model incorporating Inception module is proposed. The multiscale feature extraction network is incorporated into the generator of CycleGAN. And the residual network is combined to expand the perceptual field. Finally, the fused feature maps are reconstructed and restored to fog-free images by image restoration network. Compared with existing dehazing networks, the proposed network in this paper is more natural in the processing of image dehazing and has better results in image details as well as colors.

Heterogeneous systems consisting of multiple CPUs and GPUs are increasingly common as platforms for highperformance computing. OpenCL1 is widely used on this platform because of its cross-platform features and program portability. However, how to map OpenCL kernels onto the heterogeneous system in the presence of contention (i.e. multiple kernels compete for the computing resource) remains an outstanding problem. This is crucial to improve the efficiency of task execution. In this paper, we propose an efficient OpenCL task scheduling framework which schedules multiple kernels from multiple programs on CPUs-GPUs heterogeneous platforms. Our scheduling framework schedules kernels based on how well they match the actual running state of the current device. We show that the kernel execution is affected as the load increases. And we develop a novel model that schedule the kernel based on static and dynamic information about the kernel and the device. The framework provides adaptive and intelligent OpenCL multi-task scheduling on CPUs-GPUs heterogeneous platforms. We experimentally verified the efficiency of the framework. In the presence of resource competition, our approach achieves speedups of 1.47 and 1.61, compared to the two common scheduling strategies.

With the rapid development of e-commerce, online shopping, and the Internet of Things, regional mixed warehousing has replaced single small warehousing, not only becoming the first choice of various transportation companies, but also the mainstream configuration of various malls and supermarkets. Constant temperature is an indicator that must be achieved for basic storage, however, a large-scale storage system also poses great challenges to the local heating network. The adaptive machine learning algorithm proposed in this paper can quickly and efficiently generate a model of a large-area warehouse heating network, accurately simulating heat demand.

Salary prediction can get the income range during a certain period, which is widely used in a credit decision, career choice, and HR strategy. The decision tree is a common method for salary prediction, and it can sum up the experience of training data. However, the high dimension and high variance in splitting data remain a challenge of the decision tree in salary prediction. To overcome the above challenge, in this paper, we use Random Forest (RF) to predict salary and improve accuracy. Firstly, we preprocess the dataset by dealing with missing data, categorical data, and deleting the useless data. Then, we construct RF to eliminate the variance of a single decision tree by repeating grouping and splitting data. To decrease variance produced by multiple variables, we arbitrarily choose subsets of given variables to reduce factors. Besides, the repetition of tree model constructions in RF eliminates clashes caused by splitting data in each tree level. To verify the effectiveness of the proposed method, we compare it with other state-of-art baselines, including decision tree, logistic regression, naïve bayes, and k-nearest neighbor on the Adult dataset. The experimental results demonstrate that the proposed method outperforms related benchmarks in predicting salary.

Software system is a complex system with manual participation. Only by fully understanding the evolution law of software system can we give sufficient technical support to software network. This paper mainly discusses the structure, type, model and characteristics of software system. The digital network model of software system is comprehensively constructed, and the evolution law of software system is comprehensively analyzed from the aspects of network scale, quality and structure by using complex computer theory. In the aspect of software system construction, we should vigorously develop the development of network application system and construct the network application software system with component structure as the core. Provide good technical support for software network development and maintenance.

Understanding long-term (e.g., one year) traffic characteristics of cellular base stations (BSs) is of great value to network operators. However, there are rarely modeling results about them. In this paper, we characterize the long-term (i.e., one year) traffic patterns of thousands of BSs in a large-scale cellular network of China. We first find that the traffic distribution among BSs is highly skewed and BSs’ traffic varies dramatically in a year. In order to cluster meaningful BS traffic patterns, we use a new clustering method, in which a BS’s monthly traffic is represented by its rank in the BS’s traffic time series. In this method, we find that the thousands of BSs have six typical traffic patterns, and the patterns are interpretable: they are clearly related to two important events in China: 1) Spring Festival when a lot of people return hometown to reunion with family, 2) Double 11 Shopping Festival when a lot of people shop online. They are also related to the BSs’ geographic location and address information. Our measurement and analysis results provide useful information for cellular network providers to understand and plan their networks, and our clustering method can be applied in similar traffic pattern mining problems.

Hybrid vehicles, especially parallel hybrid vehicles, are a great scheme to improve the carbon emission condition which better suits the Chinese economy. The purpose of the article is to discover the feasibility of hybrid vehicles on the Chinese transportation net. The article used the data of the Chinese transportation net and the total amount of gas stations to analyze the huge cost of setting and integrating the same amount of charging stations. The article also compares the image of the specific structure of both serial and parallel hybrid vehicles to illustrate the feasible energy efficiency. Finally, the article concluded that the benefits of hybrid vehicles such as lower carbon emission, lower cost on popularizing electric vehicles and longer traveling distance make hybrid vehicles a potential form for Chinese transportation networks.

With the wide application of cloud computing technology and the rise of container technology, many business architectures have undergone huge changes. In recent years, container technology, microservices and other applications have gradually matured, and many enterprises' business systems have gradually migrated to the microservices model, with many microservices running in their respective business systems. In the production environment, logs play an extremely important role, and the logs in the production environment are required for business system operation monitoring, performance optimization, and troubleshooting. In the traditional business model, most of the services running in the production environment adopt independent localized log storage mode, and the operation and maintenance personnel are often unable to locate them quickly, while the large amount of log storage also puts great pressure on the storage of IT systems, and it is difficult to tap the data value of logs with the independent log mode. This paper proposes an Elastic Stack-based massive log management platform based on the business in our campus network as a prototype, deploys log collection components in various business systems in the campus network, classifies logs with appropriate rules, realizes centralized management and visual output, solves the massive log management problem of the current business systems in the campus network, and helps operation and maintenance and R and D.

In recent years, with the global warming aggravating the melting of polar glaciers, the strategic position of the polar region has been significantly improved. Improving the perception of polar environmental information has become a key factor for countries to take the initiative in the development of polar regions. Wireless sensor networks provide a good solution to the problem of polar environmental information perception. This paper describes an energy balanced wireless sensor network algorithm for polar environment information perception, namely EEB-LEACH algorithm. On the basis of LEACH, the algorithm comprehensively considers the effects of node residual energy, node distance from base station and node density within the optimal cluster radius on cluster head election, and adopts multi-hop forwarding strategy in inter-cluster communication. The simulation results show that the EEB-LEACH algorithm can make the energy consumption of nodes more balanced, delay the death time of nodes, and prolong the network life cycle.

The new generation of launch vehicle has the characteristics of more complex system functions and more complicated interfaces, so it is very important to fully and comprehensively test on the ground. At present, the monitoring parameters of some equipment in the front end of the test site still rely on manual observation for judgment, and the test efficiency and effectiveness as well as the safety of personnel need to be improved. This paper uses computer vision technology for real-time monitoring and intelligent recognition of instruments and meters, using target recognition technology based on Faster R-CNN and image processing technology based on OpenCV+PyTorvh to effectively judge the status and data of the instruments. The identification results can be transmitted to the back-end for storage and interpretation in real time through the existing test network, so that the test problems can be traced and reproduced. According to the test, the system's recognition accuracy rate reaches 100%. Compared with manual monitoring readings, the data record is more complete and the recognition efficiency is higher. The system is able to improve the test efficiency, ensure the test reliability, and provide a solution for unattended rocket test.

This paper mainly analyzes the fitting and precision of software reliability growth model based on polynomial regression model. Firstly, the software reliability testing workload and software reliability modeling framework are proposed. On this basis, the deformation software workload function and modeling function are introduced into the modeling framework. A new software reliability growth model is established. At the same time, Harry uses the framework to analyze and address the core issues in the modeling process. Based on the software reliability model parameters and two groups of actual effect data, the model framework is used to establish the most suitable growth model. Thus, the reliability of software system is guaranteed, and the integration and development of software and polynomial regression model are further increased.

In this paper, the basic principle of LDA （Latent Dirichlet Allocation）Algorithm is studied and the posterior probability distribution of topics in the document θ_m and the posterior probability distribution of words in the topic φ_zm,n is deduced from the Dirichlet distribution with parameter alpha and beta. And the parameters of LDA model are generated by Gibbs sampling results θ and φ。 Then, the accuracy of the final subject words under different iteration times is compared, and the whole keyword extraction algorithm process based on LDA is realized in Python. Finally, the results are visually displayed with keyword cloud.

Following American Satellite Navigation Global Positioning System (GPS), European Galilean System and Russian GLONASS System, BeiDou Navigation Satellite System (BDS) has become the fourth navigation satellite system in the world, and research on the BeiDou Positioning System become more and more in-depth. Gauss-Newton Iterative Least Squares (GNILS), which is now generally used in pseudo range single point positioning, has a strong dependence on the initial value and massive iterative calculation. In this paper, a PVT algorithm based on weighted least square is applied to calculate the positioning accuracy of the second generation of BDS (BD2). Measured and simulated results show that compared with direct calculation and least square, the proposed algorithm greatly improves the positioning accuracy, and has a good computational stability, thus to meet the application of civil navigation.

TSN(Time-Sensitive Networking) can provide deterministic switching services for critical traffic, which is an important development trend of onboard switching. Traffic planning is the key to onboard TSN applications. Its function is to map critical traffic to specific time slots in the network for transmission. Different planning results will cause different buffer requirements. FPGA-based TSN switch customization is an important way for onboard switch design. Minimizing switch buffer resources is of great significance to reducing FPGA size and power consumption. Based on the backtracking idea, this paper proposes a planning algorithm, named BOPA (Buffer size Optimized Planning Algorithm), to minimize the buffer size of the TSN switch. Simulation results show that BOPA can obtain the minimum value of TSN switch buffer resources according to the scene requirements under the premise of determining the topology and load.

In classroom teaching, teachers pay attention to each student's emotional changes and learning status to regulate teaching and learning to effectively improve the quality of teaching. However, the current classroom has problems such as teachers' lack of energy and delayed teaching feedback, which, to a certain extent, affect the improvement of teaching quality and hinder students' development. In recent years, with the rapid development and widespread application of information technology, new technologies such as image processing and artificial intelligence have brought new ideas and methods for research on improving teaching quality. We propose a face detection algorithm based on Yolov5, which detects the left and right head turn, head up or head down and facial expression by face images; classifies the head posture according to the left and right head turn and head up or head down, and then judges its concentration degree by combining with the detection classification of facial expression. Based on this algorithm, the concentration of classroom and meeting participants can be effectively evaluated based on face detection results, which improves the quality of classroom and meeting.

With the development of urbanization, pedestrian detection has become an important link in road traffic. Some existing pedestrian detection algorithm models are too large and the detection speed is too slow, while the accuracy of lightweight pedestrian detection can’t meet the detection requirements. Therefore, a detection algorithm that can meet the accuracy and real-time is designed. In this paper, SSD is selected as the basic model, MobileNetV2 is used as the backbone network, and deconvolution multi-scale feature fusion is added to the backbone network, ECA-Net high-efficiency channel attention module is added to the feature extraction network, and HFF multi-hierarchy structure is added to the auxiliary network to improve the detection ability of the model. In addition to ensuring the lightweight of the model, the inspection effect of the model can be improved. The experimental results show that the model designed in this paper obtains 78.6% mAP on Caltech dataset, the model size is 28M. Compared to the original model and other lightweight SSD models with higher accuracy, faster and smaller models.

Demotion blur has always been a classic problem in computer vision. In the past ten years, algorithm research in the field of deblurring can be divided into two categories. One is the non-blind deblurring algorithm based on the calculation of the blur kernel, and the other is the use of neural networks in the absence of information about the blur kernel and cam-era movement under the condition that the blur kernel is unknown. Remove motion blur. For this reason, Kupyn et al. [2] proposed a blind deblurring algorithm based on DeblurGAN. The algorithm can achieve good results in most scenes, but the deblurring effect on blurred objects with smaller scales is not obvious. The details are not prominent enough, and the grid effect is easy to produce. For this reason, this paper modifies its network structure on the basis of DeblurGAN and adds a residual module [3] as its backbone network. This paper uses Inception-ResNet-v2 [9] to extract features at different scales. Then FPN [4] is used for feature fusion, the smaller-scale feature pictures are up-sampled, and then the larger-scale pictures are convolved with the 1*1 size convolution kernel, and finally feature fusion is performed. The traditional multiscale pyramid generates features of different scales on images of different scales, and then predicts the features of different scales separately. The advantage is that the features at different depths of the network are merged, which improves the accuracy. The disadvantage is that the calculation cost is high. The advantage of FPN is that it connects the feature map from top to bottom and reduces the output of feature calculation. The advantage of this is that it can obtain more semantic information of the high-level network without losing the detailed information of the picture. Speed up the training speed and ensure the richness of feature extraction.

With the continuous development of Internet technology, how to maintain network security has become an important issue. In recent year, the number of distributed denial-of-service (DDoS) attack has been rapidly grown. Hackers take advantage of the vulnerability in the network protocol to carry out DDoS attack to paralyze the server, causing massive economic losses. Therefore, it is significant for enterprise to accurately predict and defence against malicious DDoS attacks. This paper focuses on studying a type of reflection-based attack DDoS attack portmapper, which use a third-party remote procedure call (RPC) service to launch the attack. For the experiment, the principal components analysis (PCA) and recursive feature elimination (RFE) method are used to select the most essential features from the traffic packet. Based on the selected feature, this paper uses logistic regression and support vector machine (SVM) to do the prediction. The accuracy of both methods almost reach 95% as the input training dataset increases, which proves that these two method can perform well in predicting portmapper DDoS attack.

With the gradual advancement of modernization, electric energy plays an increasingly important role in social construction and residents' life. In order to ensure the economic interests of the electric power department and consumers, regular accurate measurement and detection of electric energy in the substation is particularly important. This paper presents an electric energy metering and detection algorithm based on Stacking algorithm.

In remote sensing image processing, the rapid monitoring of urban illegal buildings or illegal occupations by remote sensing image change detection is efficient for obtaining urban geographic spatial information. In addition, the change detection has become a hot topic in the field of remote sensing, and has been widely used in various fields such as medicine, military and geographic information. Therefore, it is important to study the change detection algorithm based on remote sensing image registration. During the image acquisition by small unmanned aerial vehicles, due to the influence of the aircraft posture, height, speed and other factors, a series of problems can result in the acquired remote sensing image, such as extrusion, distortion, stretching, offset and overlap with respect to the interest targets. The key steps of change detection are image registration, whose performance can affect the results of change detection. This paper processes the image through the optimization of the low-altitude remote sensing image registration algorithm to obtain specific information of the building distribution. In addition, the paper analyzes the difference information for the low-altitude remote sensing image that has been registered. The experimental test results prove that the method is robust.

Common object detection algorithms have too many backgrounds, duplicated boxes, and a high miss ratio in detecting rotated objects, which limit their applications on the industrial site. To address these problems, this paper proposed an improved YOLOv5 to detect rotated objects. First, this paper used the K-means clustering algorithm to develop clustering analysis for trained datasets to confirm more proper anchor boxes to reduce the training time. Then this paper transferred the angle issue to a classification problem. This paper also learned angles in the original loss function combined with the circular smooth label (CSL) algorithm, thus avoiding the periodicity of angle regression. Last, this paper selected one from different detection results of an object to increase the accuracy of the detection results. The experiment showed that the proposed algorithm had a higher detection precision than other methods in the public dataset DOTA. When the proposed algorithm detected rotated objects in the dataset collected on the industrial site, its mAP reached 94.35%. This value was 8.20% higher than that of YOLOv5, satisfying the detection requirement on the industrial site.

For Unmanned Aerial Vehicle (UAV) satellite navigation and positioning system of data acquisition. This paper designs a data analysis plan for BD/GPS navigation system using FPAG as the platform. The design includes a serial receiving module and a data analysis module. The data format is based on the standard NMEA-0183 protocol. The data analysis module judges the data cycle and uses a comma count value to obtain the required navigation information and performs navigation information Code conversion until the completion of the analysis of BD/GPS data. Finally, the simulation test shows that this design of the data analysis module can be well extracted BD/GPS satellite data in the corresponding location information.

As an important part of new energy, the importance of solar energy is beyond doubt. However, due to the characteristics of photovoltaic power generation, most photovoltaic power stations are located in remote places, so it is difficult for personnel to check and clean them regularly. Aiming at the situation that the types of fallen leaves on photovoltaic panels are complex and difficult to clean up, an orthogonal algorithm for photovoltaic production abnormal data based on deep learning is proposed, and the model network method for quickly detecting the shelter of fallen leaves on photovoltaic panels and determining the shelter position is discussed. In this paper, an improved PP-YOLO is proposed to identify and classify the occlusion of photovoltaic modules. The original YOLO and the improved PP-YOLO are used to carry out the target detection experiment on the collected field data set of photovoltaic power station. The results show that the improved algorithm is effective, and the detection accuracy and recall rate are up to 96.3% and 94.2%. This method can provide technical support for real-time and accurate detection of photovoltaic panel intelligent cleaning.

In the age of big data, online shopping is now overgrowing and becoming an emerging trend among customers. Understanding user behaviors can let e-commerce platforms identify target customers more effectively and provide guests with more interest. For the fact that as the time customer stayed on the interface increases, the possibility that he or she would buy the product increases. Besides, buyers tend to browse details and comments more carefully. This paper proposed a method to predict user decisions based on user behavior. After collection and rebuilding, data were gotten from one product page. This paper uses them for training the Logistic Regression Model. After several iterations, optimal solutions can be obtained using the steepest descent method, and user behavior can be predicted. This paper uses the F1 to evaluate the model by combining the confusion matrix. Our method opens a new route to analyze and predict whether users will achieve specific behavior. It can be extended to more areas to perform more functions, like indicating whether the user is on schedule repayment.

Aiming at the tag collision problem in the actual application of the radio frequency identification system (RFID), this paper proposes an active radio frequency identification method and system supporting parallel anti-collision. The active tags work in parallel and the conversion time of adjacent active tags is partially overlapped. Significantly parallel anti-collision system shorten the identification time of the entire system, while ensuring that only one active tag sends data to the reader at the same time, which can effectively improve the anti-collision performance of the radio frequency identification system. It makes the averaged delay 5.83% of that of the traditional dual-channel system. The more tags, the more obvious the superiority of the parallel anti-collision system. The system has good engineering practical value and can be used in applications that have strict requirements on reading speed.

Isogeometric analysis is a new numerical method that has been developed in recent years based on the finite element method. The method avoids the geometric errors of the finite element method by unifying the models used for modeling and analysis, and has a Cn continuous nature. This paper studies the representation method of the nonuniform rational basis spline (NURBS) geometry used in isogeometric analysis; the stiffness matrix is derived according to basic finite element theory, boundary and constraint conditions are imposed, and the static analysis results are finally obtained by solving the linear equations. We use a two-dimensional plate with holes as an example to establish a model to enable comparison of the calculation results obtained from isogeometric analysis and finite element analysis from the perspectives of both calculation accuracy and calculation efficiency. The numerical example results show that, when compared with traditional finite element analysis, the isogeometric analysis method based on NURBS shows obvious advantages in both performing calculations and obtaining a solution. Isogeometric analysis can obtain accurate solutions with fewer system degrees of freedom and offers good flexibility and practicability.

Aiming at the problem of identifying crops in the natural environment, there will be mutual occlusion between crops and high similarity between crops, resulting in low model recognition accuracy. A multi crop recognition algorithm based on improved Mask R-CNN is proposed. The algorithm uses ResNest as the backbone network to improve the feature extraction ability, introduces Soft NMS algorithm to add confidence conditions to reduce crop missed detection and improve the segmentation accuracy, and introduces online hard example mining (ohem) algorithm to solve the imbalance between positive and negative samples, by increasing the training times of difficult samples, the model has better robustness. The experimental results show that the mAP of multi crop recognition in complex environment is 86.2%, which is 4.5% higher than the traditional algorithm.

An object tracking algorithm based on adaptive particle filtering and deep correlation multi-model is proposed to solve the problem of large numbers of particles, as well as the defects in the generation of object model in the conventional correlation particle filter. The proposed algorithm generates multi-object model by applying different adjustment rates to each high likelihood particle, and updates and predicts the particles adaptively according to the weight of correlation response graph and particle position. The proposed algorithm can adaptively adjust the number of particles according to the complexity of the tracking scene to obtain more useful particles, solve the problem of the conventional algorithm in model generation, and improve the tracking performance. The experimental results compared with some existing tracking algorithms on OTB100 datasets show that the proposed algorithm can track the object more accurately and stably under the influence of various challenging factors.

Large ships have a large deck area, many on-board equipment and devices require attitude information. if every device that requires attitude information is equipped with an attitude reference, Firstly, it will cause a waste of resources. Secondly, due to the influence of the deformation of large ships, the attitude reference of each device will be inconsistent, and then the collaborative work of large ships will be affected. Based on the attitude transfer theory, this paper uses genetic algorithm to optimize the layout of the attitude benchmarks of large ships according to the requirements of the shipboard equipment for attitude information. Simulation calculations show that this method can effectively reduce the number of benchmark layouts.

With the development of big data technology and the arrival of the information age, people's production, life and learning have undergone many changes. "Multidimensional data analysis" as a very convenient and efficient analysis mode, has been widely popularized. Computer technology can effectively promote the correct construction of occupational literacy model.

As an instrument widely used in substations, pointer meters are mainly used to detect the working status of substation equipment, so regular calibration is very important. In order to solve the problems of extraction difficulty, large positioning error and poor recognition accuracy in the reading recognition of pointer meters, a method for automatic reading detection and recognition of pointer meters based on deep learning is proposed. Use the improved EAST algorithm to extract character features, and perform module detection on the results. It is concluded that the automatic detection and recognition of analog meters under complex backgrounds has good accuracy and stability, which can meet the application needs of substations.

Distributed Denial of Service (DDoS) attacks are malicious cyber-attacks that overwhelm the target server with traffic and cause the online service unavailable. With the wide-scale rollout of Internet of Things (IoT), the DDoS attack has threatened almost all walks of life, including business and government. The accurate prediction and early prevention of DDoS attacks are necessary. In this paper, two machine learning models, the Logistic Regression model and linear Support Vector Machine (SVM), are introduced to make the DDoS attack prediction. Two dimensionality reduction methods, Principal Component Analysis (PCA) and Recursive Feature Elimination (RFE) are tried in the data preprocessing. The mean cross-validation accuracy is used to evaluate models’ performance. Experimental results indicate that the linear SVM model performs better than the Logistic Regression model on the DDoS Evaluation Dataset (CIC-DDoS2019). Besides, compared with accuracies of models using RFE, accuracies of models with PCA are higher and more stable. Overfitting is likely to occur in learning models with RFE, according to our observations of losses on the training set and the testing set.

Because the deployment position of mobile radar can not be obtained or is not accurate enough, when the recursive least squares (RLS) algorithm is used for radar position estimation based on the position of a known detected target and radar measurement data, the colored noise in the measurement data leads to the uncertainty of radar measurement noise covariance matrix, resulting in deviation or even divergence of estimation results. In this paper, a recursive least squares algorithm for exponential weighted covariance estimation based on estimation error feedback to adjust forgetting factor and adaptive window is proposed. Firstly, the nonlinear observation equation is constructed according to the radar detection model. Then, aiming at the influence of colored noise on the measurement noise covariance in the radar system, a measurement noise covariance estimation algorithm is designed, and the algorithm flow for radar position estimation is given. The simulation results show that the proposed method can change the forgetting factor and adaptive window according to the estimation error, and update the measurement noise covariance matrix in real time, so that the recursive convergence speed is faster, and the estimation accuracy is higher than that of the RLS algorithm and the RLS algorithm using residual-based adaptive estimation (RAE) for measurement noise estimation; At the same time, the algorithm eliminates the influence of colored noise in the measurement data, the estimation result of radar position can quickly converge to the real value, and the plane estimation accuracy can reach meter level.

The Hash function targeted at the short message processing is proposed in this paper. Various randomized components are designed based on the introduced thoughts of randomization, and the chaotic encryption algorithm of the variable parameters is applied to improve the avalanche performance of the Hash function. At last, the LWE problem is introduced into the operation, so to accomplish the design of the lattice-based collision resistant Hash function. Meanwhile, relative experiments are carried out, which prove the satisfactory randomization, collision resistant, and diffusion performances of the Hash function proposed in this paper.

This paper proposes an intelligent automatic train operation (ATO) algorithm based on double Q-learning to solve the problem of metro vehicles traction and braking difference and aging in order to ensure the accuracy of metro vehicles stropping to achieve the alignment between vehicles and platform doors which is required in most metro lines.

Log anomaly detection based on deep learning is one of the research hotspots in the field of computer security. It is foreseeable that the mimicry theory proposed by Academician Wu Jiangxing will further improve the detection capabilities of deep learning models, but will also bring high resource consumption and difficulty in application. Therefore, this paper proposes a mimic model construction method that uses the output of complex models as prior knowledge to train lightweight heterogeneous execution bodies and then integrates them. Finally, it is based on DPCNN and TextCNN as complex models and lightweight executions respectively. The experiment of the body structure mimic model proves that while reducing the number of parameters from millions to thousands, its detection accuracy and F1 value are only about 2% and 4% lower than the original model, which greatly retains the original model. The detection capability.

For large-scale cloud computing platform in an industrial environment, production data reading efficiency is essential, so data items are stored in different nodes of geographical area. Lack of efficient data synchronization algorithms may reduce productivity. This paper introduces a new data distribution algorithm inspired by the Susceptible-Infective- Susceptible (SIS) model to disseminate replicas under the probability restriction of reading the newest data. The experimental results demonstrate that the proposed algorithm can support reliable storage with lower latency and system complexity.

Alarm flooding is one of the important problems affecting industrial safety in modern industry. Finding similar alarm sequences and clustering the disordered alarm flood sequences by a reasonable pattern matching method is an effective way to realize alarm rationalization. In this paper, the element in an alarm flood sequence is represented by an alarm tag and a priority level. Based on this representation approach, we improved the Smith-Waterman algorithm by modified the score strategy based on alarm priorities and proposed a similarity score-based sequence segmentation method for reducing the computation cost. The effectiveness of our method was demonstrated by analyzing alarm data from an actual refinery diesel hydrogenation unit.

Maximum power point tracking (MPPT) technology can significantly improve and optimize the power generation efficiency in solar photovoltaic power generation system. The perturb and observe method is one of the most common MPPT application methods. This paper introduces an improved variable step size MPPT algorithm which has three adjustable degrees of freedom. A buck circuit simulation model is built and simulation results show that the algorithm improves tracking speed and steady-state performance.

This paper proposes a gesture recognition method based on frequency modulated continuous wave radar. The radar echo signal reflected when the hand moves is analyzed and processed in the time and frequency domain to extract the multidimensional parameters such as Doppler, range, and horizontal angle of the gesture. These parameters are spliced into a Range-Doppler-Time-Map and Range-Horizontal-Angle-Time-Map according to a fixed frame length to form a data set containing six kinds of gestures. Finally, a shallow convolutional neural network with only 5 layers is used to recognize and classify six gestures. The experimental results show that, compared with the traditional single parameter features such as Range-Time features and Doppler-Time features, the combined features of Range-Doppler-Time and Range-HorizontalTime can more accurately deal with the six gestures.

With the development of artificial intelligence (AI), data intensive algorithms, like Deep Neural Networks (DNNs), need power-consumed less but faster edge processors. In-Memory-Computing (IMC) is a promising candidate to break through von Neumann bottleneck. SRAM-based IMC provides stable, fast and low-power arithmetic ability. This article reviews the background of SRAM-based IMC, and introduces the essential influence factors and trends from the perspective of device, circuit and architecture.

Part causes down syndrome or all the third copy of chromosome 21, has been discovered for over a hundred years. Although many studies have been carried in this field to find out how to cure the patients or animals, no researchers can fully treat this disease. Thus, this research manages to analyze the expression level of the proteins encoded by the genes in mice with down syndrome by using the binary logistic regression method. The consequence shows that two significant proteins are affected most -- ITSN1_N and BRAF_N. Behind this, the predicted data are carried by 10-folds crossvalidation, then find out the result has high accuracy. Moreover, the data are randomly divided into 20%, 40%, 60%, and 80% to test the relationship between data quantity and accuracy, and it shows that the more data amount, the more precise it can be.

Due to the continuous deepening of power system reform, the continuous emergence of new businesses, and the continuous development of information and communication technology, the traditional data sharing service level of power companies can no longer meet the needs of users In order to integrate the data application business of our country's power grid companies, use the value of data can quickly improve the level of data service. We can support the function of data service by creating a data center. This article first explains the related concepts of data center, and then constructs the business architecture of power grid enterprise data center from both technical support and operation support Finally, it introduces the typical application scenarios of data center and investigates its application effects. The survey results show that the data center can integrate the island data of power companies, enhance the company's data service capabilities, and provide support for business decision-making.

Under the dual carbon goals of carbon neutrality and emission peak, accelerating the development of new energy has become a major trend in China. Distributed photovoltaic has become the fastest-growing new energy development method in recent years. In the actual power generation process, due to sensor errors, failures and other reasons, the data collected from photovoltaic power plants usually contain some abnormal data. Abnormal operation data is not conducive to photovoltaic power generation grid connection, dispatching, power generation prediction and other business operations. This article uses photovoltaic power generation as a scenario, proposes a photovoltaic power generation abnormal data detection method based on Isolation Forest and Density Clustering. It can be used when the sample set is small and the sample set is non-convex. This method does not need to rely on empirical judgment to achieve good detection results. It is very suitable for the detection of photovoltaic abnormal data.

Traditional classification algorithms tend to cause minority classes to be misclassified when classifying imbalanced data sets. In this paper, we propose an over-sampling and under-sampling algorithm based on WK-means clustering. Our method firstly uses WK-means to cluster the whole datasets, then oversamples the datasets in some regions with a large number of minority class according to the imbalance ratio by different weights of each cluster, avoids the generation of noise and effectively overcomes imbalances between and within classes. Finally, undersampling the clusters with a large number of minority class to balance the sample number of the whole datasets. The experimental results obtained from 11 datasets show that the proposed method is superior to other methods under different classifiers and evaluation criteria.

The proportionate normalized least mean square (PNLMS) algorithm is used in sparse system identification for its simplicity and adaptively step-size adjusting scheme. However, the PNLMS has the noisy input and the non-Gaussian output noise problem. A bias Compensated PNLMF algorithm (called BCPNLMF) for identifying sparse system has been proposed to solve aforementioned issues. The BCPNLMF algorithm which takes advantage of the bias compensated and the proportionate scheme, can achieve better steady-state accuracy and faster convergence speed besides identify the system parameters in noisy input and output with non-Gaussian character environments. Simulation results carried out in sparse system identification confirm the remarkable performance of the BCPNLMF, compared with other well-known algorithms.

When the traditional LT code is used for encoding and decoding, if there is no encoded packets with degree 1 in the remaining encoded groups, the decoding will stop and the decoding cannot be continued. That will affects the decoding efficiency. For this reason, this paper proposes an improved enhanced decoding method. When BP decoding stops, the remaining coded packets will be decoded using the secondary decoding method, which can improves the success rate of decoding.

Firstly, deep learning has been successful in many fields in recent years. As a core application in natural language processing, dialogue system has gradually developed from traditional machine learning to deep learning. Secondly, the development of dialogue systems is briefly summarized, and then the typical dialogue systems based on deep learning are introduced in detail. In addition, we also introduce two developer-oriented dialogue system development platforms, including PyOpenDial and DeepPavlov. Finally, we discuss the development prospects and challenges of the dialogue system, including robustness, personalized, and startup problem. We believe that this review can help researchers better understand the development status in this field.

In December 2019, a new virus called COVID-19 broke out, and in 2020, it rapidly spread all over the world. The fast rate of the spread of the virus and high mortality have brought severe harm to the health of people and the economy of almost all countries around the world. Therefore, the virus has become the object of much researches. As the study moving on, treatment and vaccine have become the leading research directions at present. For treatment, measures should be taken to protect the most severe patients to reduce the death rate, and thus we are supposed to find patients with more serious illnesses. The decision tree and Xgboost are used to get the mathematical model about protease (an essential index in judging the severity of the disease) and realize the visualization of protease data. For vaccine, we solve the problem of predicting COVID-19 Vaccination Progress in the world in 2021 using the ARIMA model, which is obtained through the mean of time-series. Eventually, we got 10-day and 3-month vaccination forecasts.

In response to the fact that most traditional communication interference recognition algorithms stay at a shallow learning level and cannot provide a detailed portrayal of the feature information inside the data, this paper proposes a deep convolutional neural network (CNN) based communication interference signal classification and recognition method to achieve the classification and recognition of five types of interference signals. This paper firstly introduces the network structure of CNN, the role of each layer, the convolution principle and common pooling operations, and then, describes the process of CNN-based communication interference signal classification and recognition, and verifies that the CNNbased communication interference signal classification and recognition method has better interference signal recognition rate and robustness through simulation analysis.

Because of the sensitivity of traditional hash algorithms to original data, it is not suitable for image content authentication. In this paper, we analyze and compare image hashing algorithms, realized a difference-value hashing algorithm (dHash), and apply the dHash algorithm to a blockchain-based image copyright protection system. dHash algorithm determines the similarity of images to ensure the uniqueness of image copyright.The test shows that the dHash algorithm has high accuracy and faster speed of images with Gaussian noise.

In order to alleviate the pressure of urban road traffic, the coordinated control strategy of regional traffic signals is studied. Taking the maximum capacity of the traffic control sub-area as the optimization objective, the regional signal coordinated control model is constructed, and the particle swarm optimization algorithm is used to solve the problem. The performance of the particle swarm algorithm is improved by adding interference items, decreasing inertia weight and adjusting learning factors. The above model and algorithm are verified by the traffic flow data of the actual road network. The experimental results show that the improved particle swarm algorithm has faster convergence speed and higher solution accuracy, and the proposed signal control model can effectively improve the overall traffic capacity of the region.