This PDF file contains the front matter associated with SPIE Proceedings Volume 12454 including the Title Page, Copyright information, Table of Contents, and Conference Committee Page.

The transmission line is a necessary way for the power generation system to transmit electric energy, and it is also a necessary condition for the configuration of the power grid. With the advancement of modern power system construction and the continuous growth of power demand, the requirements for the reliability of transmission lines are getting higher and higher, and its safe operation is related to the reliability of the power system. The purpose of this paper is to study the 3D design of overhead transmission lines based on oblique photogrammetry technology. A set of 3D construction methods based on oblique photogrammetry oblique photogrammetry technology is proposed. According to the existing experience and achievements, the method of each key link is compared, and a feasible and efficient method for constructing an overhead transmission line model based on oblique photogrammetry image sequence is summarized. The automation of key steps in the model building method is achieved through an oblique photogrammetric modeling software system. Based on the construction method of power tower nesting modeling and wire fitting modeling proposed in this paper, the three-dimensional model of overhead transmission line is designed through the example of power tower and wire modeling design, and its accuracy is verified.

Digital twin technology has been widely disseminated and gradually applied in the fields of aircraft manufacturing, CNC machining and workshop assembly. This paper mainly aims at the problems of weak mapping ability and low precision in mapping the real scene of digital twin technology into virtual space, as well as low efficiency and difficulty of visualization in virtual space, using the powerful graphics library and rendering engine of the Threejs framework, combined with the stereoscopic vision of binocular vision. Positioning technology, using Threejs as a platform to build a visual digital twin application framework, and develop a digital twin visual decision-making system. Combined with the application scenarios of electric power intelligent emergency repairs, aiming at the urgent digital transformation problems of live working robots and robotic arms in the power grid industry, Threejs technology is used to realize the construction of digital twin electric power emergency repair robotic arms and real scenes in virtual space, combined with binocular stereo vision technology Realize the acquisition of the three-dimensional coordinates required for power emergency repairs, simulate the whole process of power emergency repairs in real time, and provide new ideas for improving power work efficiency, reducing operating costs, and ensuring personnel safety.

The digital media interactive installation art takes the computer software and hardware as the platform, forms a four-dimensional space-time through the installation media, introduces the action and participants into the simulated virtual environment, and interacts with the art works by stimulating the participants' hearing, vision, touch and smell. With the development of the digital age, digital media interactive installation art design has become an important direction in the field of art design. Especially today, when green design is advocated, it is necessary to introduce the concepts of green, energy saving, ecology and environmental protection into the digital media interactive installation art design, so as to enhance the core value of installation art design. This paper discusses the art design principles, design elements and design applications of digital media interactive devices based on green design.

In view of the limitations of the current teaching quality evaluation methods of teachers in various colleges and universities, this paper at first takes five teachers in our institute as an example to establish a scientific evaluation index system according to the standards of teacher teaching quality evaluation. After the weight of each indicator is determined through the hierarchical analysis method, it overcomes the subjectivity of directly giving weights to a certain extent. Then by using the grey correlation projection method to evaluate this teaching quality indicators of the five teachers were comprehensively evaluated and sorted by the evaluation results.

The traditional wavelength division multiplexing communication network for power internet of things has the shortcomings of coarse-grain spectrum division mode, which cannot satisfy the differentiated communication requirements of flexible and diverse power services. This paper proposes an electric-elastic optical networks (eEONs) architecture that utilizes orthogonal frequency division multiplexing to achieve smaller granularity of communication bandwidth scheduling, and proposes an upper confidence bound (UCB)-based automatic routing selection algorithm. The simulation results show that the proposed algorithm can analyze the differentiated requirements of different types of power services in digital twin network, and select the most suitable routing and frequency slot.

Photoelectric tracking systems are widely used in the fields of reconnaissance, communication, and measurement. Due to different application scenarios, photoelectric tracking systems often face various interference problems such as complex backgrounds, similar targets, blurred camera motion, poor lighting, and strong mobility of moving targets, which have become a huge challenge for photoelectric detection systems. In order to improve the performance of the optoelectronic tracking system as a whole, this paper studies the monitoring and tracking technology of the optoelectronic system based on the deep neural network. This paper proposes a diagonal network and peak response regularization technology for the task of photoelectric detection target detection. The diagonal network can effectively detect the data, and the peak response regularization method regularizes the eigenvalues in the deep neural network. Plug and play is combined with common target detection tasks, which can effectively improve the effect of target detection and tracking without increasing the amount of calculation. The validation results on public datasets and UAV datasets show that the related technologies have achieved good results in photoelectric detection and tracking tasks such as human pose detection and image classification.

In order to design a practical 4-DOF robot arm, in this paper, firstly, its model was created and assembled in SOLIDWORKS; and then the model was imported to Simulink for simulation. Finally, the DH parameter method was used to construct the robot kinematics model, and he homogeneous transformation matrix between the robot arm structure coordinate system was used to obtain the positive motion and inverse kinematics solutions. The Robots Toolbox under MATLAB was used for simulation; and the third, fifth, and seventh-order polynomial algorithm was used for trajectory planning in the joint space coordinate system of the manipulator; and the fitting curve of the angle, angular velocity and angular acceleration of each joint were obtained. The results show that among three methods, the seventh-order polynomial algorithm shows the best. The robot arm’s grasping operation trajectory is smooth, and the speed, acceleration, and jerk of each joint trajectory remain continuous, which can meet the actual application requirements and provide corresponding support for the robot arm’s target grasping motion planning.

The explosive growth of data volume and dimension poses a huge challenge on service carrying capacity of power communication network with limited spectrum resources. The combination of software defined network (SDN) and elastic optical network (EON) in power communication network, i.e., EON-assisted power communication network, provides a feasible solution. To alleviate the waste of spectrum caused by non-uniform service orchestration in EON-assisted power communication network under incomplete information, this paper proposes a machine learning-based intelligent service orchestration algorithm for EON-assisted power communication network. The proposed algorithm dynamically learns spectrum allocation based on local historical information to minimize network overhead. Simulation results demonstrate that the proposed algorithm has superior performances in network overhead and selection probability of optimal arm.

for ground truths in different sizes can effectively improve the detection performance. This idea can be transplanted to classification tasks. Based on the inspiration of object detection, we re-examine the multi-scale information, simplify the steps of multi-scale fusion, and further propose two strategies for multi-scale information fusion based on weight redistribution. On classification tasks, our method maintains a certain degree of competitiveness with other finely designed fusion strategies, and improves the performance of the baseline by 1-7% on 6 datasets. Open code in: https://github.com/Clichong/StageCNN

In this paper, combine spatial clustering and Gaussian mixture models to propose an efficient and robust point set alignment scheme. This work aims to balance recall, accuracy, robustness and operational efficiency. First, the point set alignment problem is transformed into a spatial clustering problem with outliers. The main idea is to adaptively cluster the assumed matches into multiple motion-consistent clusters and one outlier cluster. The Gaussian mixture model is then fitted to each of the multiple kinematic consistency clusters. In order to pursue a higher alignment, the mixing coefficients of the Gaussian mixture model are assigned values using descriptors. This scheme is suitable for alignment problems with complex point sets. In addition, it offers better robustness and efficiency compared to existing related schemes. Analysis of the combined performance and simulation results demonstrates the effectiveness and robustness of the scheme.

At present, there are many researches and application fields of Chinese text classification, such as news classification, emotion classification, and spam identification and so on [1]. However, there are relatively few researches on the classification of Chinese abstracts of postgraduate dissertations. The abstract of postgraduate dissertation is a highly summary of the purpose, method, innovation, key and difficult points and results of the whole dissertation’s research. To recognize the abstracts of postgraduate dissertations in the library collection, and then to classify the dissertations in the library collection, is beneficial to the majority of teachers and students to consult the dissertations efficiently, and can better recommend the dissertations to readers according to their research background. Therefore, how to automatically classify postgraduate dissertations in library collection is an urgent problem to deal with. Assume that the abstract of postgraduate dissertation is written strictly according to the writing requirements, therefore, the abstract of the dissertation can represent the classification of the research field for the whole dissertation. Extract the abstracts of the dissertations as characteristic data X, label data Y was made based on the information of the school where the authors of the dissertations work in, based on the postgraduate dissertations collected in the library of the university. The dissertations studied were divided into eight categories, So Y can take integer values from one to eight. After preprocessing, the word vector and classifier were trained by the way of machine learning [2]. The results show that the classifiers can output the classification more accurately when a new abstract is randomly input to the models that has been trained.

The Appearance-based and motion-based dual-stream anomaly detection algorithm uses optical flow to extract motion features in normal video, but optical flow only considers the pixel displacement relationship between two adjacent frames, while ignoring the latent semantic association of continuous changes in multi-frame images. This paper proposes an incremental prediction-based video anomaly detection algorithm (IP-VAD), which first uses an encoder to extract corresponding features of consecutive frames, and then reconstructs feature maps from appearance and motion directions, respectively. In the appearance direction, the last frame is reconstructed by the appearance decoder; in the motion direction, the feature differences of successive frames are stacked and fed into the motion decoder, which then predicts the increment of future and last frames. Finally, future frames are obtained using the reconstructed last frame and the predicted increment. During the test, the peak signal-to-noise ratio is used to measure the reconstruction quality of future frames, and the threshold is used to determine whether it is abnormal. The experimental results on the public datasets UCSD Ped2 and CUHK Avenue show that the AUC indicators of our method are improved by 1.0% and 0.2%, respectively.

Broadband photodetectors have attractive application prospects in optical communication, image sensing, and environmental monitoring. Though narrow band gap semiconductors with good stability and low cost is an ideal candidate for achieving a relatively wide spectral response, heterojunction structures are usually proposed for constructing high-performance photodetectors with a broader spectral response and more fascinating optical-electrical characteristics. Here, a CdS/Sb₂Se₃ heterojunction photodetector was fabricated by magnetron sputtering, and morphology, microstructure, and photoelectric detection performance were studied. It is revealed that a high quality CdS and Sb₂Se₃ film with good crystallinity were gained by utilizing our growth parameters. Moreover, a CdS surface consisting of homogeneous nano islands was observed. In particular, CdS/Sb₂Se₃ heterojunction exhibits an ultra-wide photoelectric response range (365 nm ~ 1064 nm) and an ultra-fast photoresponse (< 0.2 ms). Our results proposed that these improvements on photoelectric detection performance of heterojunction is attributed to the microstructure of deposited films. This study provides a new prototype for the applications of broadband photodetector.

With the development of the Internet industry and the rise of mobile payment, the risk prevention of credit card fraud has become more difficult. With the continuous expansion of the number of cards issued, the total amount of credit, and the transaction volume, the problem of credit card fraud has become increasingly prominent. This seriously affects the normal operation of financial institutions such as banks, threatens the property safety of users, and even directly threatens the normal operation of society. In this paper, after standardizing the data set, the SMOTE oversampling method is used for sample expansion and synthesis, so that the black and white samples are balanced, and the integrated algorithm is used to train a high-performance credit card fraud detection model. Through experimental analysis, the accuracy of the model is over 99.4%, and compared with the conventional classification model, the model has superior performance.

The arrival of the 5G era provides new ideas and directions for the development of university libraries. How to promote the construction of university smart libraries has become a key concern. Based on this, based on the characteristics of users' needs of university libraries in the 5G era, this paper discusses the technical points of the construction of university smart libraries in the 5G environment from many angles, and looks forward to the future development of university smart libraries, so as to provide reference for related work in the future.

To further improve the forecast accuracy of the geometric error of CNC machine tool, a low order Fourier polynomials fitting model (LFPFM) parametric modeling method based on Improved Gray Wolf Optimization(IGWO) algorithm is proposed. Straight-axis geometric error term is analyzed, and the introduction of piecewise attenuation factor can be adjusted to the GWO algorithm is Improved, to improve the scouting speed of the Optimization algorithm and precision, the error data in all parameters based on the IGWO algorithm to find the best fitting values, and get the error about low-order Fourier polynomial fitting polynomial, Use laser interferometer to measure the location errors of three straight line axes in the CNC machine tool, and parametric modeling was carried out. The experimental results show that the machine tool error parameterized model based on the LFPFM the IGWO algorithm can fit the error data well, and the residual error is smaller, the accuracy is higher, the fitting effect is better, so it is more suitable for the prediction and compensation model of CNC machine tool geometric error.

With the consumption of global resources, photovoltaic technology has attracted worldwide attention as a clean energy source. Perovskite solar cells are recognized as one of the most promising new photovoltaic technologies because of their low cost and high efficiency. In the preparation of perovskite thin films by resistance thermal vacuum evaporator, the temperature change of the evaporation source is nonlinear, time-varying, and hysteresis. The traditional control methods have a slow response and large fluctuation, which cannot meet the precise and fast temperature control and affect the film quality. This paper proposes a fuzzy self-tuning PID evaporation source temperature control system based on Whale Optimization Algorithm (WOA). The WOA is combined with the segmentation control system together to form the WOA optimized control system. Compared with the common segmentation control algorithm, the improved algorithm reduces the response time by 16.1% and the regulation time by 23.2% while the overshoot is reduced by 75%, and the system can be stabilized with a better control effect even when the evaporation source model parameters are changed. The results show that the improved control system can improve the control effect of the system and reduce the response time and overshoot. The results show that the improved control system has effectively increased the efficiency of the vacuum coating.

Battery management technology is one of the key technologies of new energy vehicles. This paper designs the hardware of battery management system (BMS) for new energy vehicles, and uses radial basis function (RBF) neural network to estimate the state of charge (SOC) of batteries. The results show that using RBF neural network algorithm to estimate SOC can avoid the process of modeling the complex electrochemical reaction inside the battery, and can achieve high accuracy.

The mortar projectiles often land on the back of the hillside when fighting in mountainous areas. The projectile hits the target with a small landing angle, and there will be a phenomenon of rubbing the ground, resulting in a low hit rate. In this paper, a multi-rigid-body dynamics virtual prototype model is established for different targets struck by a different falling angle of the mortar projectile, the displacement and velocity data of the firing-pin and detonator as well as the incomplete firing results are obtained through simulation, the firing probability is evaluated by applying the Monte Carlo method, the contribution of the variables affecting the target value is obtained by applying the design of experiment (DOE), and the relevant parameters are optimized to realize fully fire when the parameters are in their new value range.

The paper investigates the temperature behavior and molten pool characters in the pulsed-selective laser melting (SLM) process. A three-dimensional finite element model is developed to simulate a multi-track single-layer SLM process. This model considers the temperature-dependent material properties which includes thermal conductivity, density. The simulated process consists of the heating, melting, phase change, solidification, and cooling phenomena in the powder bed. The SLM moving laser beam can be expressed as a surface moving Gaussian heat source. The numerical simulation results show that the laser power will increase the maximum temperature of the powder bed and increase the re-melting times of the powder. Step time and expose time are also affecting the maximum temperature and the re-melting times as given a constant laser power. The influence of different parameters on molten pool dimension follows the following order: laser power > step time> expose time. The temperature evolution and molten pool characters predicted by this model can be used to provide guidance for pulse-SLM process parameter optimization.

The diversity of motor controllers in terms of drive methods, feedback acquisition, and control strategies leads engineers to waste a lot of time and effort in building the underlying interfaces and functional module interactions for mechatronic system design. The FPGA design of general control of the stepper motor based on the factory method pattern splits the general control functions into relatively independent sub-functions such as communication expansion, task scheduling, feedback acquisition, control expansion, on/off control, loop distribution, and drive binding. It maps functional requirements to each sub-factory class according to the design thinking of dependency inversion, realizing the decoupling of basic control attributes and complex control strategies. Therefore, it speeds up the software development and improves the portability of the program.

Based on Unity3d virtual engine and 3ds Max modeling software, this paper takes intelligent manufacturing as the foundation and virtual simulation system as the guide to complete the construction of intelligent manufacturing simulation system in colleges and universities. Aiming at the problems existing in the practical teaching of intelligent manufacturing in higher vocational colleges, the system will take advantage of the application advantages of information technology such as virtual simulation, 3D digital modeling, Web3D visual interaction, and make highly virtual simulation modeling design of real intelligent manufacturing production lines, industrial robots and electrical control equipment. It simultaneously supports the functions of robot offline programming, virtual debugging of electrical control, virtual operation simulation, and fully simulates the real production process of intelligent manufacturing. It greatly facilitates students' in-depth understanding and mastery of intelligent manufacturing engineering planning, process analysis, engineering verification, and also provides a necessary auxiliary role for college teachers to carry out practical teaching activities of intelligent manufacturing. The design and application of the system not only effectively promoted the reform of the teaching mode of intelligent manufacturing specialty in colleges and universities, but also made a useful attempt to cultivate high-quality skilled talents of intelligent manufacturing specialty urgently needed by the society.

Capacitor is the core component of power electronic system and one of the most easily failed power electronic components. Its operating state directly affects the stability of power electronic system. In order to find and replace the faulty capacitor in time, a capacitor parameter identification method based on capacitor discharge law is proposed. Through the real-time sampling of the voltage and current curve during the charging process of the capacitor, and combining with the relationship between the voltage and current during the discharge process, the calculation model of the capacitor is established. Finally, the pre charging circuit based on capacitor is verified by experiments, and a low precision current calculation and correction method is proposed. The experimental results show that this method can accurately measure the capacitance of the capacitor, which verifies the feasibility of the proposed method.

The existing hydrogen oxygen fuel cell test system directly discharges the reaction residual gas into the external environment, which not only causes gas waste, but also has certain potential safety hazards. The fuel cell test system designed in this paper reintroduces excess hydrogen or oxygen into the fuel cell to participate in the reaction, so as to improve the utilization efficiency of hydrogen and oxygen; At the same time, the test system can also adjust the proportion of wet gas flowing through the humidifier and dry gas bypassing the humidifier in the supply gas according to the target humidity and circulating gas flow, so as to realize the rapid and accurate adjustment of gas humidity.

At present, in the assembly of casings with interference fit rabbet, there is a common phenomenon that the design tolerance and machining accuracy of each group of casings are the same, but the actual assembly accuracy of the casings is different. This is due to the small differences in the geometrical error of the interference fit rabbet within the allowable tolerance range. The coupling effect of these errors directly affects the assembly accuracy of the casings and even affects the service life of the casings. In this paper, six influencing factors are selected to deal with this phenomenon. The degree of influence of each factor on the assembly accuracy is analyzed using the finite element analysis method based on the error model. The research objects are selected as a combustor casing and a compressor casing of a type of turbofan engine, and the two casings are connected by an interference fit rabbet. The evaluation index of assembly accuracy is selected as assembly coaxiality. These six influencing factors are selected as follows: rabbet coaxiality error, rabbet interference, rabbet fitting length error, flange surface perpendicularity error, flange surface flatness error and bolt pretightening force. Based on the orthogonal test method, the sensitivity values of each factor are calculated, and the degree of influence of each factor is ranked.

The estimation of battery SOC is the core problem to be solved in the design of BMS. The accuracy of SOC estimation will directly affect the control decision of battery management system, and then affect the service life and efficiency of battery pack. In this paper, the relationship between internal resistance and SOC is analyzed by electrochemical impedance spectroscopy algorithm. It is found that the ohmic resistance RΩ, cathodic polarization resistance R2 and the total internal resistance of the battery in the equivalent circuit decrease with the increase of SOC, while there is no clear corresponding relationship between other element parameters and SOC in the equivalent circuit.

Corrugated cutter head is the key component of grassland root cutting machine, which directly affects the operation performance of root cutting machine and the improvement effect of degraded grassland. EDEM was used to establish the discrete element simulation model of soil cutting root of corrugated cutter head. The soil disturbance, root cutting resistance and stress-strain characteristics of corrugated cutter head during root cutting process were analyzed. The simulation results show that the root cutting performance of corrugated cutter head meets the requirements of grassland improvement agriculture, and the corrugated cutter head meets the mechanical requirements of strength and stiffness. The research results can provide reference for the design of adaptability, reliability and durability of corrugated cutter head.

The weight distribution of each response in the multi-response comprehensive decision-making problem of photocurable parts is highly subjective. In this paper, from the perspective of objective weighting, the entropy weight method is used to determine the weight of each evaluation index. Then, a multi-response comprehensive evaluation method based on the combination of the entropy weight method and the grey relational analysis method is proposed. Orthogonal experiments were designed with the parameters of layer thickness, laser power, over-curing and scanning spacing, and the entropy weighted grey correlation degree was sorted and analysed for the three response data of electrical energy consumption, tensile strength and surface roughness. The results show that when the printing layer thickness is 0.09mm, the laser power is 310mW, the over-curing thickness is 0.23mm, and the scanning distance is 0.09mm, the entropy weight grey correlation value of the part is the largest. The comprehensive evaluation of the part is better than other schemes. This method provides a reference for the selection of process parameters, the weight distribution of each index and the comprehensive decision-making of the additive manufacturing.

The electromagnetic railgun has the obvious advantages of high precision, wide strike range, low launch cost, and highspeed launch. However, the current concentration phenomenon of traditional electromagnetic railguns seriously reduces the service life of railguns. In this paper, a new type of electromagnetic railgun is designed in view of the shortcomings of the current railgun. At the same time, the orthogonal experiment method is used to optimize the model to select the basic type of railgun. And by using Ansys Maxwell simulation software to compare and analyze the concave, convex and planar contact modes, draw conclusions. The results show that the convex electromagnetic railgun has the advantages of uniform current distribution, large electromagnetic thrust and stable structure.

In drug research and development, in order to save time and cost, the method of establishing compound activity prediction model is usually used to screen potential active compounds, In order to become a candidate drug, a compound not only needs to have good biological activity, but also needs to have good pharmacokinetic properties and safety in human body, which is collectively known as ADMET. This paper adopts data mining technology, Firstly, the use of random forest to find the main variables of modeling is studied, and its independence is verified by high correlation filtering. The 20 main operating variables selected are MDEC-23, maxHsOH etc; Secondly, a five layer BP neural network is used to establish a compound bioactivity prediction model, which can predict the IC50 value and the corresponding pIC50 value of the compound; Then the improved BP neural network model is used to establish the classification prediction model of compounds Caco-2, CYP3A4, ERG, hob and Mn. The algorithm verifies that the accuracy of CYP3A4 is 94.3%, and the accuracy of the five models is more than or close to 90%, which is more practical than the prediction value of the improved BP neural network; Finally, the main variables of genetic algorithm are used to make the compound pair inhibit er α The value range of biological activity is optimized, which has certain practical significance.

Matrix factorization-based recommender system is in effect an angle preserving dimensionality reduction technique. Since the frequency of items follows power-law distribution, most vectors in the original dimension of user feature vectors and item feature vectors lie on the same hyperplane. However, it is very difficult to reconstruct the embeddings in the original dimension analytically, so we reformulate the original angle preserving dimensionality reduction problem into a distance preserving dimensionality reduction problem. We show that the geometric shape of input data of recommender system in its original higher dimension are distributed on co-centric circles with interesting properties, and design a paraboloid-based matrix factorization named ParaMat to solve the recommendation problem. In the experiment section, we compare our algorithm with 8 other algorithms and prove our new method is the most fair algorithm compared with modern day recommender systems such as ZeroMat and DotMat Hybrid.

In industrial production, energy is mostly provided from the combustion of fossil fuels. While the flue gas emitted after fuel combustion contains sulfur dioxide, adding a desulfurization device after the combustion unit can effectively reduce the sulfur dioxide emission. In this paper, the effects of flue gas inlet concentration, pH value of absorbent solution, and over-air volume per unit length of the desulfurization and dust removal device on the desulfurization efficiency are systematically investigated by building an experimental platform. By measuring the sulfur content in the inlet and outlet flue gases, the degree of influence of each influencing factor on the flue gas desulfurization efficiency is derived, which provides a certain basis for improving the desulfurization efficiency of desulfurization equipment in industrial production.

Aiming at the problems of locating and identifying impediment for minor targets as well as unpromising detection accuracy, a traffic sign detection algorithm based on improved YOLOv5 is propounded in this paper. First, the Addition of the Coordinate Attention Mechanism module behind the backbone network enables the model to effectively capture the relationship between location information and channels, thereby more accurately locates the area of interest. Then a detection layer is introduced into the network structure, and the feature map is continued to be upsampled after the 17th layer of the original algorithm network structure, so that the feature map continues to expand, and a feature map of 160 *160 size is obtained at the 19th layer and the feature map of the second layer of the backbone network is spliced and integrated. This results in a larger feature map for the detection of small targets. Finally, α -CIoU loss is used to replace the original CIoU positioning loss function in the model training stage to obtain higher quality anchor frame. Experiments are performed on the TT100K dataset, compared to the original YOLOv5 algorithm, the proposed algorithm improves accuracy by 3.9%, recall rate by 5.5% and mAP by 4.6% under the premise of less speed reduction, which still meets the requirements of real-time detection.

To realize the autonomous and precise take-off and landing of unmanned aerial vehicles (UAVs) on unmanned surface vehicles (USVs) and to reduce the space occupied by the embedded system of UAVs, we design a UAV-assisted autonomous landing system that is equipped with sealed charging functions for multi-rotor UAVs. Using the unmanned surface vehicle as the calculation center, the system makes use of the onboard stabilization platform to compensate for the irregular motion of the deck autonomously, which helps simplify the landing process. The system will then automatically transport the UAV into the ship cabin after the landing process has been finished, ensuring a safe environment for the UAV. In this paper, we introduce several technologies, including real-time kinematic (RTK) GPS, stabilized platform, and UAV attitude measurement and prediction, and describes how to design and implement the system. This paper has provided technical support for precise landing and automated tracking for cooperative UAVUSVs, and it also offers flexible ideas for other cross-domain collaborative systems.

Because industrial robots have uneven positioning accuracy in the overall working space, they have greatly limited their applications in the fields of high -precision production and processing. Regarding this status quo, this article focuses on analyzing the optimal working space of industrial robots, and screening and limiting the optimal working scope of the selected industrial robots to ensure that the robot end is at a relatively high position with high positioning accuracy. This article selects the collaborative robot AUBO E3 for modeling, analysis and simulation. Calculate the robot based on the Jacques matrix and the number of conditions under the spiral theory, quantitatively depict the exercise performance of the robot. The excellent working space is visually simulated, and finally uses the API laser tracker to measure the end error of the robot. As a result, the accuracy of the end point in the optimal working space is higher than the endpoint accuracy of the overall work space.

The lower limb exoskeleton robot is a wearable device that enhances the movement ability of human lower extremities. Locomotion mode recognition is an important prerequisite for controlling the lower limb exoskeleton robot. This proposed paper demonstrates a method of locomotion mode recognition of the lower limb exoskeleton robot based on residual networks. This article studies five common locomotion modes, including level ground walking, stair ascending, stair descending, ramp ascending and ramp descending. We collect locomotion data from 5 subjects using a set of 4 inertial sensors for 5 lower limb locomotion modes of different specificities. The residual network based hierarchical classifier is trained to classify the modes into a specified label hierarchy. The residual network can automatically learn mixed features without the participation of experts, avoiding manual feature extraction. It is noticeable to solve the degradation question and alleviate the problem of gradient disappearance. Some comparative experiments carried out on the selection of hyperparameters applied to optimize the residual network structure proposed. The results show that the average recognition rate reaches 96.95%, which has a prominent recognition effect on locomotion mode recognition.

An innovative curling-like amphibious robot was developed for the 2022 Beijing Winter Olympics ground and underwater torch relay. The shape design of imitation curling was applied to the robot in response to the theme of Winter Olympics. The amphibious robot had ground mode and water mode. In ground mode omnidirectional wheels drove the robot to rotate and move forward. The method of constant angular velocity rotation superimposed forward motion overcame ground slip. The nonlinear control of the system was improved by applying this actuation method when the robot slipped. In water mode the X-layout horizontal thrusters and vertical thrusters were applied to sail and hover. The position, depth, attitude and heading of the robot were precisely controlled when the amphibious robot dived into water for torch docking, which was realized through DVL, heading and attitude reference system. The multiple PD controllers with large control damping and phase margin were employed for the robot's control algorithm, so the control system was not diverge when the torch was collided or squeezed. Ultimately, the first underwater torch relay between robots in Olympic history was completed at the Beijing 2022 Winter Olympics.

Many signal processing applications simultaneously involve detecting targets' states and estimating unknown parameters. It is not been given adequate attention to joint detection and estimation (JDE) due to the differences between the evaluation metrics for detection and estimation. In this paper, we investigate the problem of JDE from the perspective of information theory. Based on the complex Gaussian scattering module of the target, we derive the joint channel for the target state and the time delay. The expression for the a posteriori probability density function is derived from the a priori statistical properties of the target state and the time delay. Furthermore, we propose the joint information as a new criterion to evaluate the JDE performance, which is the mutual information between the received signal and the target state and the time delay. The joint information is theoretical and dependent on any specific sensor. The simulation results confirm the effectiveness of the proposed joint information criterion.

In order to further improve the comprehensive picking efficiency of picking robot in China and reduce the labor intensity of manual work, the optimization research on the execution system of current picking robot was carried. Taking the working control principle of picking robot as the breakthrough point, the fuzzy control algorithm was selected as the optimization core. On the basis of reasonable allocation of weights, the fuzzy control algorithm model of the picking execution system was built according to the characteristics of picking operation. Then an executable fuzzy control system was formed through the hardware configuration and software design, and the optimization experiment was carried out. The conclusion showed that the trajectory tracking accuracy and average obstacle avoidance accuracy of the system could be maintained above 96%, the picking success rate and comprehensive efficiency of the whole machine were improved by 9.25% and 8.94% respectively, and the system had stable operation and strong advantages, which would not only realize high-efficiency intelligent picking, but also provide reference value for similar picking or harvesting machinery equipment optimization, it would be worthy of promotion.

As one of the long-standing topics in the field of soft robotics, soft crawling robots have many outstanding design achievements. At present, most of the software crawling robots adopt an integrated design. However, there are many problems in the actual use scene of the integrated design. Since the size of the body cannot be changed, the adaptive ability of the software structure cannot be maximized. This paper proposes a modular soft climbing robot, which uses the combination of a multi-chamber end effector and a single-chamber multi-chamber composite end effector to realize the modular construction of the robot. The movement process imitates the movement pattern of the inchworm, with the function of climbing vertically upwards and downwards. In this design, the number of body segments and the length of the gripper can be changed by increasing or decreasing the motion unit. Users can change the number of body segments and limb length of the soft crawling robot according to the conditions of use, which increases the environmental adaptability of the soft crawling robot. It can provide ideas for developing the application prospects of soft crawling robots.

This paper proposes the establishment of the kinematics and dynamics of tri-joint tails and combines with the swimming characteristics of fish and cephalopods. A bionic robotic fish with multi-tailed cooperative movement is designed to achieve faster acceleration ability and better stable swimming ability. Adams simulation is used for the kinematics analysis. The results show that the structure can realize sinusoidal swing well and there is a phase difference between joints. The swing characteristics of fish are simulated.

The workspace of the dual-arm robot under the obstacle working environment is analyzed. An improved D-H coordinate system is established for the six-degree-of-freedom manipulator, and the coordinates of the end of each connecting rod are solved through the forward kinematic equation. Based on the Monte Carlo method, firstly analyze the ideal workspace without obstacles, further abstract the obstacles into a cylindrical envelope, and design a collision detection strategy to analyze the workspace that is closer to the actual work scene. The results show that the workspace analysis method using the collision detection strategy can delete a large number of collision points, which provides a basic optimization for the subsequent use of fast-expanding random tree and other manipulator path planning algorithms so that the algorithm avoids selecting invalid points during spatial sampling and reduces the number of collision points. Path search time.

In order to improve the mobility and adaptability of mobile robots in complex environments, this paper proposes a kind of connecting rod four-legged composite robot leg structure. The leg mechanism adopts the design of series and parallel composite connecting rod. Each leg contains two degrees of freedom of hip joint and knee joint, and the joint motor is installed on the fuselage. The main body of the wheeled structure is the driving wheel, which is mounted on the shank. The dead point position is used to ensure the stability of the link structure during the wheel motion. The objective function of optimal design of leg mechanism is obtained by statics analysis, the parameters of leg linkage mechanism are optimized by multi-objective optimization genetic algorithm, and the motion space of foot end is solved. Through Adams on the design of wheel leg compound mobile robot Trot gait, Walk gait and wheel leg convert motion simulation, the simulation results show that the foot type sports center of mass on the direction of gravity changes in the distance is less than 4 mm, in the process of movement joint torque biggest not more than 1.75 N ꞏ m, and a physical prototype was made for experimental verification. The simulation results and experimental results show that the designed linkage type compound leg structure can realize the basic stable movement of the foot and the stable switching function of the wheel foot.

As the robot enters the carriage or container for stacking, it is inevitable to develop a reasonable stack shape detection method to timely find the wrong situations such as stacking boxes tipping and non-conforming to packing constraints, and timely correct them to ensure the smooth stacking. Therefore, a high precision stack detection method based on three-dimensional reconstruction of solid state lidar (Livox Mid-70) is designed. The actual palletizing type point cloud was obtained by laser SLAM reconstruction. Combined with the theoretical palletizing type point cloud result, the overall root mean square error of the actual reconstructed palletizing type and the theoretical planned palletizing type was calculated by point cloud registration technology, and the palletizing effect was tested. The point cloud registration experiment is carried out by using the point cloud obtained by simulation and the point cloud extracted by planning stack shape. The results show that the stack shape detection method can meet the actual requirements in detection accuracy and detection time, and verify the effectiveness of the method.

The four-degree-of-freedom robotic arm is designed and its trajectory is planned in this paper to meet the higher requirements for arbitrariness, repeatability, and stability of the robotic arm in space position in the application of the robotic arm. A comparison of the joint angle trajectory planning effect was carried out using the linear interpolation method and the quintuple polynomial interpolation method, and the motion trajectory of the robot arm was obtained. The results show that the trajectory planning by the fifth polynomial interpolation method results in smooth motion and low vibration of the arm, and the mechanical structure of the arm is reasonably designed with correct geometric parameters, which is of high application value.

The effectiveness of Model Predictive Control (MPC) for path following systems is well known. However, the commonly used control methods usually do not consider the complex computational complexity of fast dynamic systems, which is one of the main reasons for the resulting overshoot. This technical note considers the optimization problem of the controller, calculates the optimal solution of the control time through the genetic operator, then uses the block method to reduce the control time, and finally provides the optimal error weight matrix to deal with the error transient problem , so that the controller has a shorter calculation time and less error overshoot.

In order to solve the problems such as frequent shutdown of the production line and increased material loss due to the large amount of hard caked tobacco after the moisture regain is loosened, a set of caked tobacco dividing device is designed at the entrance of the HAUNI conditioning cylinder. The test data shows that the application of the new design increases the amount of caked tobacco separation and falling into the cylinder to an average of 9 pieces, and the amount of loose hard slices is also reduced to an average of 9.8 Kg per 8000Kg production batch. The design of the device improves the production stability, reduces the loss of materials, and effectively meets the requirements of high-quality development of the enterprise.

Aiming at the low-efficiency and high-cost manual sorting status of the assembly line and the need to select a specific product among many goods, a conveyor belt sorting device based on the color, shape, label and characteristics of the goods is designed. According to the design requirements of the system, the detection principles of the Kirchhoff circle detection algorithm and the AprilTag labeling algorithm are analyzed, and the trained model is deployed to the embedded terminal based on the TensorFlow Lite parser. Finally, combined with the STC90C52 microcontroller, a new method based on the characteristics of the goods is proposed. The method of classifying with different algorithms not only improves the operation efficiency of the equipment but also reduces the sorting error rate. The experimental results show that the device can sort all kinds of goods efficiently and achieve the expected design effect.

Most of the existing coating machine need manual operation, such as power on, power off, adjusting the gap of paper feeding, oil thickness and so on. Experienced operators know how to adjust, but there will be errors, affecting the quality of products. This paper designs an automatic UV coating machine, which can solve the above problems. In this design, PLC intelligent control system is used instead of manual adjustment operation, and the feed gap and oil thickness are automatically detected by photoelectric sensor. And the system with memory function, according to different materials can adjust the parameters, the implementation of the best effect. Software intelligent operation, reduce the technical requirements of operators, so as to improve the production efficiency and quality stability of products. Solidworks is used to design the structure of the whole machine. Fluent is used to analyze the flow field of the coating process. The integrity of the machine is verified by combining theory with practice.

Driver behaviour is a major cause of traffic accidents. In response to this problem, this paper proposes a solution to increase the driver's reaction time by changing the vehicle's stability through the toe-in angle The toe angle of a vehicle is an important parameter in vehicle design, which has a great influence on the stability and steering ability of the vehicle. In this paper, the influence of the toe angle on the vehicle stability is determined by analyzing the influence of the vehicle toe angle on the lateral force generated when the vehicle wheels are deflected. This paper simulates the force of different toe angles when the wheels suddenly turn by the carmaker and establishes the relationship between the stability of the vehicle and the toe angle when the wheels are deflected by analysing the lateral force on the vehicle under different toe angles. A mathematical model to analyse the effect of toe angle on vehicle stability. The results of the modeling simulations show that the increase of the toe angle can effectively reduce the lateral force on the vehicle when the wheels are deflected. From this result, it is concluded that increasing the toe angle can effectively increase the stability of the vehicle.

The mechanical structure, driving loop structure and control system are designed for a robot that can recognize the color of cylinder and carry it to a designated place. In the mechanical mechanism part, the hand, wrist, arm, body, base and connecting parts are calculated. In the driving part, the air pressure and motor are combined, and the components in the air pressure circuit (pneumatic motor, control valve, cylinder, pneumatic actuator, etc.) are selected. The control part uses embedded raspberry pie and Open CV motion detection and color recognition system to complete the detection and resolution of handling objects, Firstly, the processed data is transmitted to K60 single chip microcomputer (SCM) by serial protocol instruction. After operation, analysis and processing, the SCM issues action instructions to each executive component, thus completing the action control of robot traveling, grasping, lifting and putting down objects. It cooperates with various sensors for information feedback to realize accurate closed-loop control of the robot. After physical assembly and debugging, the robot can realize the color recognition of the cylinder, and carry the cylinder to the designated position.

In view of the low reliability, difficulty in shifting and disengaging, and "double gear" faults of the fixed axle transmission equipped with a hydraulic power assisted shift mechanism under low temperature conditions, starting with the shift principle of the fixed axle transmission, combined with the structure and working process of the shift mechanism, this paper analyzes the technical changes of the core and key components of the hydraulic power assisted shift mechanism under low temperature conditions, combined with the above analysis, It is pointed out that the hydraulic power assisted shift mechanism should take measures such as replacing the low-temperature hydraulic oil, adding a heating device and setting a gear inspection device to solve the adverse effects caused by low temperature, so that the fixed axle transmission can work more stable and reliable. After the test verification, the shift mechanism adopting the above three measures has avoided the occurrence of common low-temperature faults to a great extent and achieved good results, It has a good reference for the use of the modified hydraulic power shift mechanism of the fixed axle transmission on heavy vehicles under low temperature conditions.

Aiming at the structure of a type of radar antenna main tower, the topology optimized weight reduction design and research were carried out by the finite element analysis software ANSYS Workbench. A three-dimensional model of the radar antenna main tower was established, and its structure was topologically optimized with the unit material density as the design variable, the mass retention percentage as the response constraint, and the minimum compliance as the optimization objective. After optimization, the mass of the radar antenna main tower is reduced by about 17.8%, while its structural static strength and modal properties remain the same as before optimization.

The front wheel kingpin caster angle of cars is the inclination angle between the axis of the kingpin and the vertical line of the ground. It is one of the key points of the technical testing that affects the safety performance of automobiles, and it is also one of the key topics of today's research. Some researchers found that the front wheel kingpin caster angle has a certain influence on the car driving. However, there is still a research gap on its influence on the trajectory of the car when turning. Therefore, based on the establishment of the bicycle model, this paper analyzes the trajectory of the car return to alignment and explores the influence of front wheel kingpin caster angle on the steering stability under the condition that there is no human intervention in the car and the car's torque is solely used to bring the car back into alignment. The research found that the front wheel kingpin caster angle has a positive effect on the time and the rate of the car returning to alignment by affecting the aligning torque. When the front wheel kingpin caster angle increases, the trajectory line changes from a curve to a straight line more quickly after the turn is over. Because of this result, the research suggests that the car's front wheel kingpin caster angle can be appropriately increased to enhance the car’s rate of returning to alignment, thereby improving the steering stability of the car.

As many enterprises and factories rely on indoor heavy equipment to complete production work, due to the limited indoor working space during the transportation and installation of this heavy equipment, it is more difficult to use traditional lifting devices or large lifting devices due to their large size. It is difficult to enter the room to complete the corresponding work, and each time it is necessary to build and dismantle the lifting device according to the site conditions, resulting in difficult construction, low efficiency, poor versatility, and other problems. Therefore, a new type of indoor installation transportation device is urgently needed to solve this problem. Using innovative design methods to design a new type of installation transportation device is an important method to solve the above problems. Given the current problems of indoor installation transportation devices and the user requirements of such devices, this paper is based on a variety of innovative design theories such as QFD, AD, and TRIZ to research, establish a device innovation design process model, and design a new type of indoor installation and transportation device based on the model, taking GIS indoor heavy equipment as an example, and verify the effectiveness of the device.

The vehicle charging port locking control system designed in this paper can prohibit the insertion of the charging gun under dangerous working conditions, prohibit the charging when the charging gun is not inserted in place, and allow the charging gun to be pulled out when the battery is fully charged or reaches the charging limit but the whole vehicle is not unlocked, which improves the charging safety and the utilization efficiency of the charging pile.

The tire is a component of the vehicle in direct contact with the pavement surface. The friction coefficient of tire and road surface is inseparable from the driving safety, mobility and economy of the vehicle. This paper presents a method based on code and formula simulation to estimate the effect of friction between each tire and the road and the driving state of the whole vehicle. By classifying different road conditions, different tires and different speeds, using the Magic formula, rolling friction torque and substituting a series of vehicle basic data, the corresponding tire-road friction coefficient is obtained. All basic vehicle parameters and tire parameters are obtained from the manufacturer's official website, instructions and other channels. Then the linear images are processed by Kalman filter. These images show that different tire types, different road conditions and different driving habits will affect the friction coefficient. The feasibility of the conclusion is verified by simulation in Adams car 2020, to put forward suggestions and improved methods for the comprehensive performance of vehicle driving from many aspects.

In view of the high failure rate of the transmission part of heavy-duty tracked vehicles and the lack of effective fault judgment means and methods, this paper puts forward a fault diagnosis method of the transmission part of heavy-duty tracked vehicles based on big data on the basis of summarizing the general structure and working principle of the transmission part, analyzing its common fault types and the proportion of various types of faults. According to the fault types, in vibration The key parts of the transmission part are monitored in terms of temperature and pressure, and the key data are collected. According to the type of data collected and the size of the data set, the fault data extraction method, computer deep learning method and data migration learning method in big data are proposed, so that the computer can quickly and accurately identify and diagnose the faults of the transmission part of the vehicle, and provide intelligence for the fault diagnosis of the transmission part of heavy tracked vehicles Efficient means and methods.

The stroller is a very widely used category of children's products, but the existing stroller has the problems of single function, short service life and the design is not humanized enough. To solve the problem of strollers and build a long-term harmonious relationship between strollers and users, this work explored the design process of multifunctional strollers based on Kansei Engineering and guided by modular design ideas, and extracted a set of design methods. Expecting to integrate more scientific and humanized design ideas and methods into the design of strollers, so as to provide strong support for the improvement of existing stroller products and design and development of new products.

Cleaning boats are frequently used to remove trash from offshore areas to combat sea pollution. With a small self-propelled rubbish cleaning boat as the model, this paper presents a design for converting solar energy into power for the ship's control system. Wireless signals are transmitted via Bluetooth modules and several components of the system are controlled by Arduino platforms. Solar panels can be automatically deployed using this device. In the event of harsh weather, the emergency protection system can be activated to ensure the reliability of the system's operation.

The real contact area of mechanical seal has a great influence on the frictional properties and sealing effectiveness of the interface. Based on fractal theory and the influence of friction, the computer model for the real contact area of dimensionless mechanical seal is established. Based on LabVIEW, a simulation program for calculating the real contact area of mechanical seal is produced. By means of the simulation program, the real contact area of dimensionless mechanical seal-interface payload ratio-interface fractal dimensions-coefficient of scale characteristics curve is obtained. The result shows that the real contact area decreases with the monotonically nonlinear-increasing of interface payload ratio, and decreases rapidly then slowly with the increasing of coefficient of scale characteristics. With the increasing of interface fractal dimensions, the real contact area increases and then decrease, that is there is a fractal dimensions which can make the real contact area reaches a maximum value.

Aiming at the lightweight design of a vehicle-mounted radar array structure, the static analysis and modal analysis of the design model were carried out by Ansys to obtain the deformation, stress distribution cloud diagram and the first 6 order natural frequencies and modes of vibration. The calculation results show that the safety factor of each part of the radar array structure is above 2 and the maximum deformation is less than 20mm after the main array plate weight is reduced by 8%. The overall stiffness and strength of the radar array structure are good, which can meet the design requirements.

With the continuous improvement of people's living standard, cars have been popularized to ordinary families. According to incomplete statistics, the number of vehicle rescues caused by automobile failures or accidents accounts for about 10% of the automobile ownership every year[1].Anchoring vehicles will lead to road congestion, especially in the rush hours of urban traffic, and even cause some sections of traffic to be paralyzed. Therefore, how to rescue the breakdown of vehicles quickly and effectively, dredge the city road is particularly important. This paper introduces a modular vehicle rescue robot concept design to supplement traditional rescue methods and improve the efficiency of vehicle rescue under congested road conditions, thus solving the problem of low rescue efficiency of traditional rescue methods under congested road conditions.

Telemedicine is likely to be an increasingly available medical mean for modern patients and those who are inconvenient to go to hospitals. As a kind of Chinese traditional rehabilitation training, eight-section brocade is more suitable for patients of most ages and they can promote physical quality. However, it is difficult for non-professional patients to complete standard posture without doctor guidance. Therefore, in this work, a method of remote data collection, posture scoring, and classification based on convolutional neural network of Siamese Network will be presented. 1, 204 images are used as template pose data and other 1, 204 pictures are patient’s poses. To train the network to have a better understanding of the image's sophistication, a Convolutional Neural Network (CNN) is used in this work, which can take an input image and apply learnable weights and biases to the image's numerous objects. Meanwhile, Siamese network is mentioned as a method to compare two vectors with different outputs so that posture classification and evolution can be realized. The average similarity of the score is 0.2172, as can be shown. The score range is -1 to 1. The closer the score is to 0, the more the patient's activities resemble the template.

With the progress of society, people's needs are changing and higher requirements are being put forward for living standards and living environments. Therefore, the use of machine vision and information technology to solve the current problem of dangerous systems is a good solution. This paper explores the design and development of a school safety warning system based on machine vision algorithms. The school safety early warning system based on machine vision is an intelligent management system that can achieve real-time monitoring and early warning functions, which is important for achieving school safety management, by analyzing the existing system requirements and drawing conclusions.

The rapid development of big data and artificial intelligence technology has been applied to various industries. Flexible employment is a new sharing mode under the sharing economy. The lack of a digital platform for tax issues between freelancers and employment enterprises is an urgent problem to be solved. This paper develops an intelligent tax planning platform based on the current situation analysis, and proposes a new tax planning algorithm based on the advanced technology of big data and artificial intelligence, which helps enterprises to carry out intelligent tax planning under the premise of legal compliance, reduce company labor costs, and increase the freelancers’ revenue. The platform and algorithm solve the contradictions and pain points between enterprises and freelancers. Secondly, this paper gives the process of tax planning and Spark ecosystem, architecture running on the Spark Cluster. Finally, the detailed algorithm and analysis process are summarized. Through the actual case analysis, compared with the existing algorithms, the intelligent tax planning platform algorithm proposed in this paper has achieved better results, helping companies reduce costs by 21.9% and increase individual revenue by 18.5%, and achieving good economic and social benefits. This research is grounded on our regional, helped hundreds of enterprises to plan taxes.

The development of electric power is of great significance to the progress and national security. Several large-scale power outages occurred in recent years have sounded the alarm for the development of my country's electric power industry. At the same time, with the continuous progress of computer and automation technology, the continuous advancement of power grid informatization and artificial intelligence construction, the construction of distribution network fault intelligent processing system is of great significance to the development of power grids and society. The purpose of this paper is to study the design of an intelligent processing system for distribution network faults based on artificial intelligence. Focusing on the distribution network fault information processing and control system, combined with the construction practice of the distribution network intelligent fault processing system, the development status, theoretical basis, system design and engineering construction of the distribution network fault intelligent processing system are studied. Based on the model, the architecture and technical framework of the fault intelligent processing system are analyzed, combined with the development status and actual needs of the power grid, seven business functions of the system are constructed, and the application effect of the distribution network fault intelligent processing system is analyzed. After running the fault intelligent processing system, the voltage qualification rate has been improved to a certain extent, which ensures the safety of the distribution network operation.

In the process of informatization construction of public sports teaching resource platform in colleges, network construction is the foundation, and resource construction is the core element of network construction. The construction of the network physical education resource platform has brought new development opportunities for school physical education, and of course it also faces a series of challenges brought by the network information society. This paper first expounds the current situation of the construction of the network public sports teaching resource platform system in colleges, and builds the construction of the university network public sports teaching resource platform system based on deep learning technology. Finally, it is concluded that online education researchers should use big data and deep learning analysis technology to Provide learners with appropriate learning resource push services, so as to better meet the learning needs of online learners, realize humanized and personalized learning, and further play the role and value of the resource platform.

With the advent of the era of big data, semi-supervised learning algorithms are becoming more and more popular and widely used. Self-training is the most widely used semi-supervised learning framework. The performance of the classifier obtained by self-training mainly depends on the selection of high-confidence samples during the self-training process. To study the effect of the base classifier on the performance of the DRNG algorithm, we use KNN, decision tree, and SVM base classifier to conduct experiments on four medical images. The experimental results show on the most datasets, the classification performance of the DRNG algorithm on the KNN (K=1) base classifier is higher than the other three base classifiers. The DRNG algorithm matches the KNN(K=1) base classifier better.

In recent years, with the rapid growth of the number of cars, the problem of "difficult parking" has become increasingly prominent, especially in the parking lot. It often takes a long time for drivers to find a free parking space. Low parking efficiency seriously affects the driver's parking feeling, which is an urgent problem to be solved. In view of this phenomenon, based on the analysis of the internal structure of the parking lot, this paper abstractly establishes the parking guidance data model, comprehensively considers the restrictive factors affecting the driver's parking psychology, then uses the improved Dijkstra algorithm to optimize the guidance data model, and finally finds out the optimal parking path, which greatly improves the parking efficiency and parking feeling.

With the development of society, people's life pressure is increasing, and the number of stroke patients is also increasing year by year. At present, stroke has become the second leading cause of death in the world. Therefore, it is necessary to predict whether a person is more likely to have a stroke. At present, studies around the world have established a variety of stroke prediction models based on genetic or pathological indicators. However, the accuracy of the prediction models derived from these studies is still insufficient. This study designed a new method to predict stroke. After the datasets are collected and processed, the SMOTE method is used to balance the datasets in order to obtain better results in the future prediction. In the stage of constructing new patient features, eight complex new features with different correlations were created. After completing the construction of new features, this study uses the machine learning model algorithm built in the Sklearn function library in Python to train and predict the processed datasets. Experimental results show that the prediction accuracy of most algorithms has reached more than 95%, which has been greatly improved compared with previous studies. Finally, this study uses PyQt5 and QtDesigner to build a simple GUI, which can be easily operated by users who have the need to predict stroke.

As one of the important means for human beings to understand and develop the ocean, the application of underwater robots in marine science and engineering has attracted more and more attention. Underwater robots are divided into two categories, one is remotely operated underwater vehicle (ROV) and the other is autonomous underwater vehicle (AUV). Autonomous submersibles do not carry cables. Its energy is installed on the robot. Its task execution process is controlled by computer. Complete the task independently according to the robot program. Because there is no cable constraint, further analysis data can be collected after returning. AUV has the characteristics of autonomous navigation and large-scale observation. This paper presents a method based on convolution neural network, which realizes the effective perception and recognition of underwater environment. Through the ROS operating system, we integrate the control and environmental perception of the underwater robot, and design an underwater robot that can avoid obstacles and recognize autonomously.

The height of Tetraena mongolica Maxim is a marker to measure the growth rate and an important index to analyze its growth condition.In this study, the Tetraena in a 50m×50m size sample plot were used as the research object, and the images taken by the UAV mounted camera and the 3D point cloud data taken by the ground-based radar were used to combine the 2D data from the UAV with the 3D data from the ground-based radar. The unsupervised classification and supervised classification by support vector machine, random forest and maximum likelihood methods are performed, and the height is obtained by kriging interpolation of the point cloud data, and the image data and point cloud data are fused. Using this method, the maximum height of 0.504m and the vegetation coverage is 13.11%, which provides a new research method for the effective protection of the Tetraena.

Due to the paucity of data for this VTuber industry and the volatility of the data itself, in recent years the processing of datasets in the VTuber industry has remained at the statistical stage, and providing accurate forecasts of industry trends has been quite challenging. In this paper, a new way of thinking about dealing with the VTuber industry on Bilibili datasets is conceived, namely using Neural Networks (NN) and machine learning algorithms to implement predictions. Bilibili is a well-known video-sharing website and also the largest live streaming platform for VTubers in China. More specifically, the prediction of datasets in the VTuber industry using the Long Short-Term Memory (LSTM) model is done based on the TensorFlow machine learning framework. In short, the dataset is processed to match the dimensions of the TensorFlow library input by dividing and preprocessing the training and test datasets, and then the prediction is implemented by the LSTM model. Finally, the error metrics are calculated by Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Square Error (RMSE). The experimental results show that the LSTM model achieves a very high accuracy match in predicting the VTuber industry dataset, achieves a more competitive performance in data processing, and can provide more effective industry predictions for those engaged in the VTuber industry.

At present, most power plants are not intelligent for workers, and there are weak safety awareness. The combination of computer vision and the actual engineering site, such as real-time monitoring of workers' protective equipment wear, can greatly reduce the occurrence of safety accidents.This paper collects a dataset of 2,035 images from a power plant video record based on the deep learning network DenseNet to train the model. On the test dataset, the average accuracy was 0.84 and the recall rate was 94%.In this paper, this paper detects the wearing equipment of workers based on computer vision, which provides help for complex field security management.

Over recent years, offline handwritten Chinese character recognition which based on convolutional neural network has become a research hotspot, and the accuracy rate is constantly improving. At the same time, there are problems such as large model size, many calculation parameters and model redundancy. In response to this, this paper proposes a lightweight and improved MobileNetV2 offline handwritten Chinese character recognition model based on attention mechanism. By adding an attention mechanism to the pooling layer of CNN, the L2 norm-constrained Softmax classification function is used in the final classification. In this paper, a research experiment is designed to compare the improved model with the traditional method and the classic CNN model. Experiments show that on the ICDAR-2013 competition test set, the accuracy of this method reaches 96.12%, and the model size is 48.34MB, which is improved in both recognition rate and volume compared with previous methods.

To reduce the casualties caused by workers not wearing safety measures in time at construction sites. For the generic target detection model with high complexity and large model, which cannot perform helmet wearing detection in real-time, a real-time helmet wearing detection algorithm based on YOLOv3 is proposed, with the mobile end network as the feature extractor of the proposed method. For helmet wearing detection is often done outdoors, so to filter the environmental noise. The channel attention module is introduced to optimize the feature extraction when multiplexing the multi-scale feature maps. Finally, to weaken the problem of inadequate gradient back propagation brought by the IOU function, the CIOU loss function is used to optimize the gradient back propagation. The experimental results show that the method in this paper can balance the accuracy and detection rate of the detection model with an average accuracy of 82.47% (mAP). Compared with the model of the original YOLOv3 network, the model size of this method is only 22.3% of the original model, and the detection rate is significantly improved compared with the original method and can meet the requirements of real-time helmet wear detection.

As an important branch in the field of computer vision, the exploration of video understanding is still ongoing, and its important part of video understanding has become one of the research hot spots. Firstly, aiming at the problems of large amount of convolution network parameters and model degradation caused by the deepening of network layers, DDensenet network is proposed, and dynamic grouping convolution module (DGC) is adopted, which greatly reduces the amount of parameters, improves the calculation efficiency and alleviates the problem of gradient disappearance. Secondly, the improved convolutional attention module (S-CBMA) is added, and the adaptive module is added to its spatial attention mechanism, which can distribute the weight of features and weaken the background at the same time. DDensenet network model achieves 95.8% and 79.8% average classification accuracy on Sports8 data set and Olympic16 data set respectively. The results show that the model has faster convergence speed and higher classification performance than C3D, ResNet3d, Two-Stream and other network models.

This study suggests an enhanced A* algorithm to address the issues of numerous traversal nodes, numerous inflection points, and huge turning angles in the path planning of the traditional A* algorithm. The traditional A* algorithm is first improved in this paper's heuristic function and search efficiency using dynamic weighted averaging; next, the path planning of the traditional A* algorithm is smoothed using Bezier curves; and finally, the effect of the improved A* algorithm is tested in the built scene using gazebo simulation software and ROS system. The outcome demonstrates that the modified A* algorithm has superior path planning capabilities compared to the traditional A* algorithm.

The accuracy of the current digital text intelligent classification results is low, which leads to the classification efficiency and quality can not achieve the desired goal and restricts the collaborative development of text information accuracy, efficiency, intelligent retrieval, and mining. To solve this problem, based on the traditional text classification method, this study introduces the concept of machine learning and designs a new intelligent classification method for digital text. Firstly, the digitized text is preprocessed to reduce the length of the digitized text; then, the feature weight of the word frequency of the digitized text is calculated to provide data support for the subsequent text intelligent classification; finally, a machine learning classification model is established to intelligently classify the digitized texts. The experimental analysis shows that the accuracy of the new text classification method is more than 96. 87% in all categories of data sets, which has a significant advantage over the traditional method, and the classification processing time is shorter and the feasibility is higher.

In order to solve the current problems of insufficient detail extraction and poor visual effect after high magnification reconstruction of face images, a super-resolution method is proposed for single images of faces based on generative adversarial networks. Channel attention is added to the generative network to extract richer facial details, and the idea of iterative up and down sampling layers in the depth inverse projection network is borrowed to make the reconstructed image with good visual effect after high magnification. For the discriminator network, the normalization layer, which would destroy the image contrast, is removed. The experimental results show that the reconstructed images are more realistic and the visual effects are improved compared with Bicubic, SRCNN, LapSRN and SRGAN.

In today's rapid development of artificial intelligence and big data, when people make decisions, they want to give guidance and advice through data-related information, such as: Will banks lose customers? Is the energy consumption data normal? How much is the price? These predictions and judgments can be built through deep learning models to obtain corresponding results. Through the research of artificial neural network in deep learning, this paper sorts out the key steps of its application, and proves through experimental data that the prediction model constructed based on artificial neural network has higher accuracy and less time-consuming.

The detection of SAR images of ship recently become one of the important research direction, but the SAR image imaging feature is not obvious, the face of a complex scenario testing result is not ideal, can produce serious leakage inspection by mistake, and so on and so forth, in addition, the deployment in the actual situation, the detection accuracy and speed cannot be both at the same time, lead to much difficulty in the practical application. This paper designed a anchor-free detection algorithm TA-FCOS on the basis of FCOS, which can effectively reduce the number of parameters. Meanwhile, we proposed an attention module TA, which was embedded into the backbone network to effectively improve the detection accuracy. Finally, we conducted experiments on the HRSID data set. AP50 and mAP reach 85.9% and 57.8%, respectively. Experimental results confirm the effectiveness of our algorithm.

In recent years, with the rapid growth of the number of cars, the problem of "difficult parking" has become increasingly prominent, especially in the parking lot. It often takes a long time for drivers to find a free parking space. Low parking efficiency seriously affects the driver's parking feeling, which is an urgent problem to be solved. In view of this phenomenon, based on the analysis of the internal structure of the parking lot, this paper abstractly establishes the parking guidance data model, comprehensively considers the restrictive factors affecting the driver's parking psychology, then uses the improved Dijkstra algorithm to optimize the guidance data model, and finally finds out the optimal parking path, which greatly improves the parking efficiency and parking feeling.

Breast cancer is a common cancer that threatens many people’s lives, especially women. Early and accurate diagnosis of breast cancer contribute much to improving life qualities of the patients. However, it is challenging to construct a reliable diagnosis model to diagnose breast cancer. In this study, to achieve better performance of the breast cancer diagnosis, a model called ANN-SVM model that hybridizes artificial neural network (ANN) and support vector machine (SVM) is proposed to categorize the breast tumours into benign or malignant. To be more specific, in the process of training artificial neural network, the hidden relationships between features and classification results are detected. The hidden layer in artificial neural network, which contains the information of these relationships can be extracted as input of support vector machine. The final classification results are given by SVM. ANN, SVM and the proposed ANN-SVM model were experimented on Breast Cancer Wisconsin (Diagnostic) Dataset. The performance of each model was evaluated based on accuracy, F1 score, and false-negative rate. From the results, the proposed ANN-SVM model reached the highest accuracy of 98.24% and F1 score of 0.9760, and it achieved the same false-negative rate of 1.17% as artificial neural network did, which was lower than the false-negative rate achieved by support vector machine. It is shown that the proposed ANNSVM model outperformed single ANN and SVM.

In the field of robotics, mobile robots have always been a hot topic of research. Aiming at the problem of mobile robot formation, multi-robot collaborative control technology is adopted, and the leader-follower algorithm is realized through the design of the leader and follower roles, and the multiple robot formation control is realized, and the dynamic window algorithm is integrated to achieve the formation to avoid obstacles. First, the leader-follower algorithm is introduced and theoretically derived. Secondly, the dynamic window algorithm added in the formation collaborative control process is introduced and theoretically derived. Finally, through the simulation of cooperative control of formations, the dynamic window algorithm and the artificial potential field algorithm were used for simulation and comparison of the formation travel process and the path trajectory traveled by the final formation, respectively.

In recent years, the extensive application of knowledge atlas in China has attracted the attention of the industry and plays an important role in knowledge production and management. By using knowledge graph technology, knowledge can be organized more orderly and organically. Combined with certain knowledge mining and analysis algorithms, users can obtain their own knowledge information more quickly and accurately, so as to realize intelligent reasoning and decisionmaking. At present, the knowledge graph has been widely used in the commercial field, and there are corresponding products in intelligent search, machine Q & A, intelligent recommendation, etc., such as the new generation search engine, Siri and other chat robots, community recommendation systems, etc. By studying the basic types of knowledge atlas technology, this paper analyzes the application of knowledge atlas in knowledge problem, recommendation system and military intelligent decision-making system, and points out the universality and universality of the application of knowledge atlas technology in modern intelligent field.

In recent years, the application of Siamese networks for visual target tracking has brought a great improvement in tracker performance, allowing both accuracy and real-time performance. However, the accuracy of Siamese network trackers is largely limited. Object tracking algorithm with multi-scale channel attention is proposed based on the Siamese object tracking algorithm of fully convolutional classification and regression. Firstly, the backbone network ResNet50 is improved, which combined with multi-scale channel attention for feature extraction and enhancement. Then, adding a spatial attention mechanism to focus on information about the location features of the target image within each channel after the features are extracted from the template branch. At last, it performs fusion, classification and regression successfully. Compared with other advanced trackers, our approach achieves higher accuracy and success rate, especially in complex scenarios such as fast motion, occlusion, similarity interference, and scale changes.

With the rise of deep learning, the end-to-end speech recognition model has received increasing attention. Currently, the performance of the end-to-end speech recognition model has been further updated on basis of on the proposal of the Conformer Framework, which has been widely used in the field of speech recognition. However, these models perform poorly on edge hardware due to large memory and computation requirements. This paper compresses the model by exploring the application of knowledge distillation in end-to-end speech recognition based on Conformer. By combining with the network structure of the model, the training loss function is redefined, and on this basis, the influence of different temperatures on the performance of the student model in knowledge distillation is explored.

This paper designs a multi-functional remote monitoring system for wards based on single chip microcomputer (SCM). Taking STM32 SCM as the core, the system detects the patient's body temperature and heart rate, and sends the detection signals to the SCM using temperature sensor and heart rate sensor, which is displayed on the display module after analysis and processing. If the patient has abnormal conditions such as fever, too fast or too slow heart rate, the system would alert the patient through buzzer alarm, and send a warning on the mini program of the medical staff by WIFI to remind the medical staff to treat in time. The liquid status in the infusion bottle is monitored in real time by the liquid level sensor. When the liquid is almost finished, not only the buzzer sounds to remind the patient to close, but also the system can send the information to the mobile phone by WIFI to show that the liquid has been finished. The nurse can add medicine or take the needle to the patient in time. According to the mini program, the inclination angle that needs to be adjusted by the hospital bed can be selected on the mobile phone, and the motor can drive the hospital bed to adjust the corresponding angle. Patients can call the monitoring center by pressing the key if they have any other needs, and can have a direct conversation with the nurse or doctor in the monitoring center. The system has the functions of human-computer interaction, simple operation, remote monitoring and so on, which has a strong practical significance for real-time understanding of patients and improving the efficiency of treatment.

This paper designs a kind of voice recognition intelligent waste classification system in order to solve the complex household waste and difficult classification. This system consists of mechanical part and control part. The mechanical part completes the design with SolidWorks 3D design software. Based on the control core of STM32 chip, the control system integrates the human body sensing, infrared distance measurement, voice recognition, voice broadcast, serial communication, man-machine interactive module and other modules. The system is able to give a prompt when the wastes are full and conduct man-machine interaction with users. It can recognize the names of the wastes by users, judge the properties of the wastes and give a prompt. The mechanical system and drive device can rotate and uncover the waste bins so as to classify the wastes intelligently. This device is of high accuracy, fast response, small volume, easy use and strong promotion value.

The prediction of the remaining useful life of machine tool spindle bearings contributes to the increase of the useful life of CNC machine tools. In order to solve the problems that traditional modelling method faces the complex process and low accuracy, a new method to predict the spindle bearing remaining useful life combining the neural network and weighted average de-nosing method is proposed in this paper. First, the method uses a convolutional neural network to extract features from the vibration signals of bearings; secondly, a gated cyclic unit is used to perform regression analysis on the extracted features to estimate the remaining useful life of the bearing; finally, the weighted average de-nosing method is used enhance the accuracy of the predictions. The effectiveness of the proposed method is verified by experimental data.

This control system is a modern new device which integrates sensing technology, computer technology, network communication technology and mobile communication technology. The system is mainly composed of STM32 single chip microcomputer (SCM) as the control core, sensor as the information detection element, with the corresponding executive mechanism, voice broadcast, display and mobile device. In addition to the functions of conventional push-key call and obstacle detection between elevator doors, it also has the functions of mobile phone remote monitoring, voice recognition control, telephone call alarm and so on. By detecting and recognizing the call signals from keys, mobile phones and voices, the SCM sends forward and reverse commands to the motor and PWM wave signals to the steering gear, so as to realize the opening and closing of elevator lifting and elevator doors. After the elevator reaches the target floor, there will be corresponding voice prompts. When the elevator door is opened, the infrared tube sensor is used to detect whether there are obstacles between the elevator doors. If so, the buzzer makes an alarm. If not, the elevator door closes with delay. All running states of the elevator will be displayed on the OLED screen in real time, and reported to the mobile phone for remote monitoring by users. The corresponding functions are realized after programming and physical debugging. The mobile phone remote call and voice call functions of the system have more prominent advantages during the COVID-19 or in hospitals, which can realize contactless riding, avoid contact with infected germs and greatly save waiting time.

The current semantic recognition methods for long and short texts have low recognition accuracy and slow recognition speed for texts with obvious length differences, which can not accurately identify the deep semantics contained in the texts and restrict the development of text data. To solve this problem, based on the traditional text semantic recognition method, the natural language processing (NLP) artificial intelligence principle is introduced, and a new design of long and short text semantic recognition method is proposed. Extract the semantic features of the original text, and use the frequency method to process the lexical features in the text by weighting. Then, obtain the feature weights of the long and short texts and use the NLP artificial intelligence method to distinguish the unknown texts and identify the semantics of the long and short texts online. The experimental analysis shows that the accuracy of the semantic recognition methods proposed in this paper is more than 96. 87%. The accuracy is higher. The recall is less than 35. 27%. The recall is lower and the recognition advantage is significant compared with the traditional methods.

In recent years, the construction industry has developed rapidly. As the largest amount of building materials, the consumption of concrete is increasing day by day. As a kind of green and healthy concrete, the development and application of recycled concrete(RC) can not only alleviate the occupation of natural resources by concrete and ensure the sustainable development of human society, but also have important significance for resource recycling and environmental protection. Based on this, this paper puts forward artificial intelligence algorithm(AIA), discusses its application in RC bridge engineering, and briefly analyzes the basic mechanical properties, processing technology and mix design of RC; The grey neural network predictive control of the durability index of RC bridge is discussed, and the network output results of the predicted and measured strength of RC bridge of AIA are measured through experiments. The results show that the predicted results have a certain guiding effect on the subsequent construction control, and verify the feasibility of applying AIA to RC bridge engineering proposed in this paper.

According to a related report, housing prices in New York will have similar price changes in the future as they do now, with minor swings and a modest increase overall. This study hopes to take a more original view, which means, less popular sides would be considered, and through machine learning to predict the trend of house prices in New York. The information of data was about house prices and real estate in New York. The first step is called conversion for data type, converting the string data in the document to the data format called INT. The second step is normalization, the values are mapped from 0-1. The proposed prediction of this article study uses three algorithm methods: Support Vector Machine (SVM), Linear regression, Neural network. And then the article will go for the hyperparameter configuration analysis. The paper concludes that practically every aspect of the data affects house prices through the comparison of the three forecasting models. Multiple linear regression models provide unstable outcomes, and the outcomes always vary from time to time. The neural network model's prediction results are relatively stable, and when the hyperparameters are kept constant, the prediction results are roughly the same. The most stable outcomes are predicted by the SVM model.

Identifying the metabolic pathways through computation is critical in metabolic engineering for the design of synthetic pathways. The constraint-based pathfinding of searching the metabolic pathway from a given starting metabolite to a target metabolite has been broadly used to find the metabolic pathways of synthesizing valuable metabolites, but few people have committed to adjusting a mixed integer linear program (MILP) model to search for the metabolic pathways from arbitrary starting metabolites to a target metabolite. In this paper, we present a novel metabolic pathway prediction method for discovering the metabolic pathway from arbitrary metabolites to a target metabolite in an adjusted metabolic network according to the feature of metabolites. This has the potential to enhance the biochemical correlation of metabolic pathways significantly. To summarize, our method is a valuable metabolic pathway prediction method for identifying alternative metabolic pathways and further improve the study of metabolic pathway prediction methods from arbitrary starting metabolites to the target metabolite.

The scientific and technological revolution centered on artificial intelligence is leading a new round of educational innovation. this paper takes the era of artificial intelligence as the background, according to the characteristics of artificial intelligence, Analyzes the opportunities and challenges faced by the practical teaching of curriculum political practice, adopts the method of distributing questionnaires for investigation, analyzes the problems existing in the current the practical teaching of curriculum ideological and political, expounds how artificial intelligence promotes the development of curriculum political practice teaching, and finally puts forward relevant policy suggestions for the reform of curriculum political practice teaching.

There have been many studies on the survival analysis of patients with non-small cell lung cancer (NSCLC). However, most of the studies are based on the extraction of tumor radiomics features based on the tumour label outlined by the physician, followed by a combination of clinical and pre-treatment PET/CT image features of the patient for survival analysis. Survival analysis of patients with locally advanced NSCLC based on whether pre- and post-treatment FDG-PET can be performed without tumors label using a deep learning approach. The consistency index (C-index) of the convolutional neural network model was 0.67 when using pre- and post-treatment FDG-PET, suggesting that simultaneous reading with pre- and post-treatment PDG-PET can predict the probability of patient risk.

Artificial intelligence(AI) is being integrated into enterprises’ strategy for improving management performance, while many managers do not know the focus of AI research, nor do they know what specific aspects AI is applicable to and the possible impact. The main purpose of this research is to analyze the change that brought by technologies based on AI on enterprise, to determine the compound effect on enterprise management reform. This paper proposes a framework for fully understanding the opportunities and challenges on enterprise management reform brought by AI to help companies break down their prejudices about AI. Through bibliometrics method and literature review, we find that AI applications can promote the beneficial changes of the three relationships in enterprise management, including the relationship between enterprises and customers, superiors and subordinates, as well as the relationship between traditional production factors and data factors. Meanwhile, enterprises may face difficulties in security and privacy, law and ethics, and culture and talent. All in all, enterprise management will embrace the opportunities of innovation in theory and practice in the era of AI if we can effectively prevent or deal with the possible problems.