In order to reduce the total overhead of the MEC (Mobile Edge Computing) system, the weighted sum of latency and energy consumption of all devices are considered as the optimization objective, and the problem of task offloading is solved in a multi-user multi-channel mobile edge computing system. Specifically, multiple user devices are able to offload computationally-heavy tasks to the MEC server over a wireless channel. Considering the difference in residual energy among multiple smart devices, an energy factor is introduced to measure the bias of smart devices between energy consumption and latency. A reinforcement learning scheme based on the Q-learning algorithm is applied to co-optimize the offloading decision, the allocation of computational resources, and the selection of wireless channels. Simulation results show that the algorithm can effectively reduce the delay and energy consumption of task processing and accommodate more users.

The channel resources of RFID ( Radio Frequency Identification) systems are limited, and when multiple tags compete for the same frequency or time slot, it can lead to collisions and conflicts. In order to optimize the communication efficiency of broadcast channels, a collision prevention algorithm for RFID broadcast channels in the Internet of Things based on frame time slot ALOHA is proposed. This method introduces the concept of frame time slots and divides the communication time into time slots; By analyzing the probability of occurrence of idle, successful identification, and collision states within the time slot, the cause of collision in the

broadcast channel is obtained. By combining Bayesian algorithm and Poisson distribution rules, the probability distribution of the number of tags is calculated to estimate the number of tags within the range of the reader and writer, and the next frame length is adjusted based on the calculation result of the number of tags. If there is still label collision problem within the adjusted frame time slot range, FastICA( Indcpendent Component Analysis) independent principal component analysis is used to transform the label recognition problem within the frame time slot into an EPC(Electronic Product Code) encoding generation problem, thereby achieving parallel recognition of multiple labels within a unified time slot and avoiding collision situations. The experiment shows that the estimation of the number of labels proposed is accurate, which can improve the label recognition rate within the time slot and effectively improve the propagation efficiency of the broadcast channel while ensuring the stability of the communication channel.

In order to suppress the influence of low-frequency magnetic field interference, such as geomagnetic diurnal variation, on the measurement results, multiple magnetometers are usually used for synchronous measurement. The time synchronization error has an obvious influence on the suppression effect. The influence of different time synchronization errors is studied on the JOM-5SF Overhauser magnetometer in geomagnetic detection and instrument sensitivity evaluation, based on the magnetometer developed in the laboratory. Two Overhauser magnetometers are used to conduct experiments on the campus of Jilin University. After comparing

the experimental results with the evaluation results of professional institutions, it is found that the smaller the time synchronization error, the smaller the difference between the magnetic field values of the two instruments, and the more accurate the sensitivity of the evaluation instruments by the synchronous method.

In order to fulfill the observation requirements for the total magnetic field intensity of geomagnetic stations, a specialized magnetometer architecture is independently designed for the stations, and the Overhauser magnetometer is developed for station monitoring. Sensitivity evaluations are conducted using both single-station direct measurement and dual-station synchronized methods under field conditions with low noise and in environments with high electromagnetic interference. The experimental results from both direct measurement and synchronized methods indicate that the sensitivity of the JOM-5J magnetometer can reach 0. 02 nT at a 1s period. It is capable of replacing the GSM-90F for applications in earthquake precursor observations and long-term

volcano monitoring.

A single embedded processor is difficult to efficiently complete the massive computing tasks such as image processing. Therefore, a set of SoC(System on Chip) with image processing function is designed based on FPGA(Field-Programmable Gate Array) and Coretx-M3 processor kernel. Based on Xilinx’s Kintex-7 FPGA and Arm’s Cortex-M3 kernel, the processor architecture is implemented on FPGA. The memory, bus system and basic peripherals are designed using IP(Internet Protocol) core and Verilog, and are connected to the processor through the bus. The image processing unit is designed, and the commonly used digital image processing

algorithm is mapped to the hardware description language. And the bus interface is designed to connect to the processor, providing the image processing capability for SoC. Based on Keil MDK tool and C language, the drivers for the peripheral and image processing unit of SoC are written, and the system function is simulated. And the digital image processing based on Matlab and the image processing unit in SoC are fully compared and tested by taking the binarization algorithm as an example. This image processing SoC has excellent performance and all the advantages of FPGA and SoC. The author has successfully developed a SoC with image processing function based on FPGA platform. The system is board-validated on Xilinx’s Kintex-7 family, model XC7K325TFFG676-2 FPGAs. This design reflects the high flexibility and efficiency of the system designed on FPGA platform, and provides a solution to solve the disadvantages of a single embedded processor that is difficult to efficiently complete the massive computing tasks such as image processing. The system is designed based on a reconfigurable platform, which can realize the customization of peripheral functions according to requirements, and has the advantage of higher flexibility.

Rolling bearing is one of the key components in rotating machinery, and long-term mechanical operation leads to wear easily. Traditional fault diagnosis relies on feature extraction, but due to loud noise during mechanical operation, effective signals are drowned. And the fault diagnosis network structure is complicated and there are too many parameters. Therefore, a bearing fault diagnosis model based on variational mode decomposition and deep learning is proposed for bearing wear detection. Firstly, the bearing signal is decomposed by VMD( Variational Mode Decomposition) and denoised by Hausdorff distance. Secondly, the

selected effective signals are inputted into the network structure of one-dimensional convolutional neural network and gate recurrent unit to complete the classification of data and realize the fault diagnosis of bearings. Compared to common bearing fault diagnosis methods, the proposed VMD-1DCNN-GRU(Variational Mode Decomposition- 1D Convolutional Neural Networks-Gate Recurrent Unit) model has the highest accuracy. The experimental results verify the feasibility of the proposed model for the effective classification of bearing faults, which has certain research significance.

Currently, the docking of single-column steel pipe towers in power systems mainly relies on manual implementation, which has a high risk factor and is time-consuming and laborious. Aiming to research the autonomous docking technology of single-column steel pipe tower for this specific technical condition and environment, a vision-based navigation-based hydraulically driven autonomous flexible docking system for single- column steel pipe tower is proposed. The single-post steel pipe tower autonomous flexible docking system uses a microcontroller that enables autonomous positioning and docking through image tracking control, and the

computational performance of the microcontroller makes it easier for technicians to operate. Numerical simulations and hardware tests are carried out for the docking of a single column steel pipe tower in the two- dimensional plane. The results show that the effectiveness of the proposed method is verified by controlling the motion of the steel pipe tower assembly at the docking interface using a small steel pipe tower model with air bearings in the two-dimensional plane with a minimum propellant thruster and a small control moment gyroscope.

UAV(Unmanned Aerial Vehicle) aerial photography, characterized by multi-angle, large field of view, and large-scale scenes, often results in images with numerous small objects, complex backgrounds, and difficult feature extraction. To address these issues, a new model, CA-NWD-YOLOV5 ( Coordinate Attention- Normalized Wasserstein Distance-You Only Look Once v5) is proposed. Based on the YOLOv5 model, a multi- scale detection layer is added to the head network to extract the features of small targets. It also incorporates a CA attention mechanism into the backbone network to prevent the model from overlooking target location

information. Lastly, the normalized Wasserstein distance loss function replaces the loss function based on intersection ratio, enhancing the model’s sensitivity to small targets. Experiments on the VisDrone2019 dataset demonstrate that, compared to the improved YOLOv5 model, the CA-NWD-YOLOv5 model can effectively enhance the detection accuracy of small and medium-sized targets in UAV aerial photography images. The mAP_ 0. 5 of the improved algorithm reaches 50% , proving its effective application to the detection of small targets in aerial photography.

In order to solve the problems of uncertainty, discontinuity, randomness and nonlinearity in stock price forecasting, a TRSSA-ELM ( Tent Random Walk Sparrow Optimization Algorithm-Extreme Learning Machine) stock price forecasting model is proposed. Firstly, adaptive Tent chaotic mapping and random walk strategy are used to improve the algorithm, which enhances the diversity and randomness of the population and improves the local and global optimization ability of the algorithm. Secondly, the performance of TRSSA( Tent Random Walk Sparrow Optimization Algorithm) is verified by using single peak, multi-peak and fixed multi-peak

test functions. Compared to SSA( Sparrow Optimization Algorithm), AO( Aquila Optimizer), POA( Pelican Optimization Algorithm) and GWO(Grey Wolf Optimizer), TRSSA algorithm has better convergence speed, accuracy and statistical properties. Finally, because the ELM ( Extreme Learning Machine) model randomly generates weights and thresholds, which reduces the prediction accuracy and generalization ability, TRSSA algorithm is applied to optimize the weights and thresholds of the ELM model, and the TRSSA-ELM model is tested in Sanan Optoelectronic stock data set. The experimental results show that TRSSA-ELM model has better prediction accuracy and stability than SSA-ELM, ELM, SVR(Support Vector Regression) and GBDT(Gradient Boosting Decision Tree).

Traditional methods for identifying table structures are difficult to fully learn complex table structures such as merge cells with multiple rows and columns, blank cells, nested cells, and are lack of information in the process of extracting features. An OCR(Optical Character Recognition) table structure identification method based on improved SLANet (Structure Location Alignment Network) is proposed. Firstly, the lightweight CPU (Central Processing Unit) convolutional neural network is used and attention mechanism is introduced to enhance the generalization ability and explanation ability of the network. The information vector obtained by training is

inputed into the lightweight high-low level feature fusion module to extract features, and then the outputted features are aligned with the structure and position information through the feature decoded module to obtain the prediction label. Experiments show that compared to EDD ( Encoder-Dual-Decoder), TableMaster and other models, the accuracy of the proposed method has been significantly improved, reaching 76. 95% , and the TEDS (Tree-Edit-Distance-based Similarity) has reached 95. 57% , which significantly enhances the model’s ability to identify complex table structures and provides an optimization strategy for identifying table structures.

Although traditional recommendation systems use graph structure information, most of them only consider the basic attributes of users and items, ignoring the important factor of contextual interaction information between users and items. Even if contextual interaction information is taken into account, there is a lack of attention in the layer combination stage. force mechanism to assign weight. To solve this problem, a CIAGCN (Context Information Attention Graph Convolutional Networks) recommendation model that integrates contextual interactive information and attention mechanism is proposed. This model utilizes the contextual interaction

information of users and items while applying the high-order connectivity theory of graphs to obtain deeper collaborative signals. An attention mechanism is introduced in the layer combination stage to improve the interpretability of this stage. The model was experimentally compared on the Yelp-OH, Yelp-NC and Amazon- Book data sets. The results showed that the model had a certain effect compared with other algorithms, indicating that the recommendation effect was better than some traditional recommendation models.

At the traditional image processing algorithms for the detection of steel surface defects, there are problems such as low recognition efficiency and a high false detection rate of leakage. The YOLOv8-DSG (Deformable Convolution Network Squeeze and Excitation Network Generalized Intersection over Union) steel surface defect detection algorithm is proposed. Based on the traditional YOLOv8 algorithm, several improvements are made. Firstly, the DCN ( Deformable Convolutional Network) is embedded in the C2f ( Convolution to Feature) module of the Backbone network, which enhances the feature extraction ability of the model under

complex background conditions. Secondly, the SE ( Squeeze and Excitation network ) attention module is introduced into the Neck network, which highlights the important feature information of the steel surface and enhances the richness of the feature fusion. Lastly, the GIOU ( Generalized Intersection Over Union) loss function is used instead of the original CIOU(Complete Intersection Over Union). Compared with CIOU, GIOU introduces the minimum enclosing frame area ratio, which can more accurately measure the overlapping area of the frames. The experimental results show that the YOLOv8-DSG algorithm achieves an average accuracy mAP of

80% on the NEU-DET dataset, which is 3. 3% higher compared to the original YOLOv8 algorithm. And it has a low rate of misdetection and omission, demonstrating higher detection accuracy and arithmetic efficiency. This algorithm can play an important role in quality inspection.

The L_OOA(An Improved Osprey Optimization Algorithm) is proposed to address the issues of the original OOA (Osprey Optimization Algorithm), which is prone to local optima and slow optimization speed. Firstly, to maintain population diversity, the Tent chaotic mapping strategy is adopted to initialize the individual positions of the population. Secondly, by introducing the Levy strategy to update the position of the Osprey, the Osprey Optimization Algorithm can improve its ability to jump out of local optima. The spiral curve strategy is introduced into the Osprey optimization algorithm to improve its computational accuracy. Finally, comparative

experiments are conducted with other intelligent algorithms on the CEC2021 ( Computational Experimental Competition 2021)testfunction set. Experiments prove that L_OOA has better accuracy and faster speed.

The warship’s common computing environment integrates computing and storage resources through virtualization technology to build a public infrastructure platform for warships. However, it is limited by the space and energy consumption requirements of the maritime combat platform. Optimizing the virtual machine deployment strategy is an important development to reduce the energy consumption level of the warship’s commoncomputing environment and improve basic resource support capabilities. Several commonly used virtual machine optimization deployment methods are compared and a virtual machine deployment strategy for warship’s common computing environments is proposed based on an improved flower pollination algorithm. An improved maximum

and minimum distance method is designed and applied to the initial population generation process to enhance the initial solution. To improve the quality, a local search strategy with an information exchange mechanism is proposed by introducing the hybrid frog leaping algorithm. And an adaptive switching probability strategy is proposed to balance global pollination and local pollination, and generate an optimized deployment plan for mapping virtual machines to servers. It is verified through simulation experiments that the proposed virtual machine deployment optimization strategy can significantly reduce the energy consumption level of the warship’s common computing environment.

When using data from a single data source to complete tasks, there may be significant errors in the data, and there may even be data missing, which can affect the progress of the task. A dual channel data association and fusion algorithm based on fuzzy mathematics theory is proposed for this purpose. The correlation of dual channel data is measured and the missing data in the dual channel data is predicted according to the missing data prediction process. The missing data in the dual channel dataset is filled in to obtain complete dual channel data. The dual channel data is standardized, and the principal component analysis is used to calculate

the similarity between the dual channel data and the principal components, obtaining the comprehensive support level of the dataset, and obtain effective data. By using fuzzy mathematics theory, effective data is fuzzified, and the closeness between the fuzzification results and real data is calculated to determine the data fusion weight, in order to achieve dual channel data association and fusion. The experimental results show that using the proposed algorithm for dual channel data association fusion, when the total number of data reaches 1 500, the value of the comprehensive evaluation index exceeds 9, indicating that the proposed algorithm can improve the accuracy of dual channel data association fusion and has good dual channel data association fusion results.

Weil pairing and Tate pairing are widely used in encryption, signature, password exchange and cryptosystem security analysis. It has been suggested that the computational efficiency of Tate pairing is better than that of Weil pairing, but this problem is still doubtful and needs to be further verified. The parameter selection algorithm of binary group structure proposed by Miller belongs to probabilistic algorithm, and the algorithm efficiency is not high. To solve the above problems, the analysis models of Tate pairing and Weil pairing on execution efficiency are established, and a new method is proposed to find the parameters of the distortion value by using the quadratic relation of the order of the elliptic curve. The research shows that when the distortion value is small, the computational efficiency of Tate pairing is better than that of Weil pairing, which is consistent with previous studies. However, when the distortion value is large, the computational efficiency of Weil pairing is better than that of Tate pairing, and the time complexity of the new method to find the distortion value parameter is less than that of Miller method O(M). Compared with Miller’s probabilistic

method, the new method is deterministic. The correctness of the analysis model is verified by analysis and example, and the new method greatly improves the efficiency and accuracy of parameter selection.

The era of big data has arrived, and the digitization of archive information is the future development trend. How to protect the privacy of digital archive information is a key research topic in the computer field. At present, the archive privacy protection algorithm based on blockchain technology has problems such as poor protection effect and long operation time. In order to solve the problems existing in traditional methods, a digital archive information privacy protection algorithm based on blockchain technology is proposed. Firstly, apply blockchain technology to the privacy protection process of digital archive information. The specific protection

process is as follows: the data owner uses symmetric encryption algorithms to encrypt the digital archive information and upload it to the private chain; At the same time, generate a secure index of digital archive information and upload it to the alliance chain; The data user generates a query threshold for the keywords to be queried, sends it to the private chain, obtains the query results on the private chain, and sends them to the alliance chain. The alliance chain cooperates with the private chain to verify the correctness of the query results. If it is correct, the alliance chain will send the converted encrypted data to the data user. The experimental results show that the privacy protection algorithm for digital archive information of the proposed method has good privacy protection effect and practical application effect.

In order to reduce the error between the target drilling trajectory and the actual drilling trajectory measurement results, a drilling trajectory measurement error correction method based on particle swarm optimization algorithm is proposed. A drilling trajectory calculation model is established, the values of drilling inclination angle, drilling orientation angle, and drilling azimuth angle are determined, and drilling trajectory data is collected. The sources of errors in drilling trajectory measurement is analyzed, an error transfer state space model is constructed, and the historical errors are merged to complete the calculation of drilling trajectory

measurement errors. The error correction objective function is constructed with the goal of minimizing the measurement error of the borehole trajectory. The velocity and position of particles are updated, and the fitness function is constructed. The objective function is solved by continuously updating and calculating the fitness function to complete the correction of the measurement error of the borehole trajectory. The experimental results show that the roll angle error in each direction of the proposed method is only between 0. 3° and 1. 8 °, and the drilling trajectory is highly fitted with the actual value curve, which can effectively correct the trajectory

measurement error and provide valuable reference for the actual exploration work of underground engineering.

The current intelligent prediction methods for traffic flow have not analyzed and considered the spatiotemporal correlation of the road network. We conduct research and improvement to address this issue by adding spatiotemporal correlation information to the intelligent prediction methods to solve the problem of reduced prediction accuracy caused by the lack of spatiotemporal information. The spatiotemporal correlation of the urban road network is analyzed by combining the map connection of the traffic network and the vehicle traffic delay. Considering the spatiotemporal correlation of urban traffic, based on the GCN( Graph Convolutional Neural)

network and LSTM ( Long Short-Term Memory) network methods, the urban traffic flow prediction method considering spatiotemporal information based on GCN and LSTM is studied. Urban traffic flow prediction network is optimized and trained by using the open source urban traffic flow dataset. The performance of LSTM, BiLSTM (Bidirectional Long Short-Term Memory) network and different number of nodes in solving the traffic flow prediction problem is compared. The results of this research show that the proposed method can effectively predict urban traffic flow, and the accuracy of the proposed method is improved compared with the prediction method without considering spatiotemporal information. This research can provide a theoretical reference for traffic prediction in intelligent transportation systems.

In view of the defects that the projection matrix learned from LatLRR ( Latent Low Rank Representation) can not explain the importance of the extracted features and preserve the local geometry of data, a novel method named LLRSP (Latent Low-Rank and Sparse Projection) with dual neighborhood preserving and feature selection is proposed. The algorithm first combines low-rank constraint and orthogonal reconstruction to hold the main energy of the original data, and then applies a row sparse constraint to the projection matrix for feature selection, which makes the features to be more compact and interpretable. Furthermore, a l^2,1 norm is introduced to regularize the error component to make the model more robust to noise. Finally, neighborhood preserving regularization is applied on the low dimensional data and low-rank representation matrix to preserve the local manifold geometrical structure of data. Datasets results of extensive experimental on various benchmark show that this method can obtain better performance than other state-of-the-art methods.