最新刊期
卷 21 ， 期 6 ， 2022
- 摘要：At present， the recommendation is mainly implemented through the user’s preference， which obtained through user’s ratings and item attribute， to find the nearest neighbor. However， these algorithms ignored the influence of item quality and time factor when mining user’s preference and ignored the user’s interests in item type when generating the recommendation candidate set. Therefore， this paper proposes a MF-based collaborative filtering algorithm combining with item quality and time weight （QTW-MFCF）. This algorithm constructs user-item type interest matrix by combining item type information and user-item ratings modified by quality factor and time factor. Then the recommendation candidate set is generated according to user’s current item type preference. Finally， the matrix factorization （MF） model is used to predict user’s ratings of all items in the candidate set， so as to get top-N recommendation list. The experimental results on two public movie datasets show that compared with the benchmark algorithm， the accuracy and recall of the algorithm are improved by 25.17% and 44% on M1 and 32.9% and 28.59% on M2 respectively. The experimental data show that the algorithm can obtain user preferences more effectively and further improve the recommendation quality.关键词：interest migration;collaborative filtering;item quality;time weight;item type;matrix factorization
- 190
- |
- 58
- |
- 0
发布时间：2022-06-24 - 摘要：In recent years， quality safety accidents of geographical indication agricultural products have been frequently reported. In order to effectively avoid and control quality safety risks， a quality safety risk assessment and early warning model of random forest （RF） and deep confidence network （DBN） was proposed. First of all， the quality safety risk assessment system is constructed by combining the five elements of total quality management （TQM）， namely "human， machine， material， method and environment"， with the four stages of planting， processing， logistics and sales. Secondly， the stochastic forest model is used to reduce the dimension of the index system， and 13 evaluation indexes are determined. Finally， DBN is used to construct the quality safety risk assessment and early warning model， and Yunnan Pu 'er tea is selected as an example for empirical study. The results show that the accuracy of RF and DBN model applied to the risk assessment of agricultural products with geographical indication can be as high as 96.67%， which is significantly better than 90% of reverse propagation neural network （BP） and 85% of support vector machine （SVM）. This result provides a reference basis for reducing the quality and safety risk of agricultural products with geographical indications and ensuring their brand influence.关键词：RF;DBN;geographical indications of agricultural products;risk assessment;early warning
- 87
- |
- 29
- |
- 0
发布时间：2022-06-24 - 摘要：When constructing the deep neuro fuzzy system with high-dimensional data， the requirements for the accuracy and time of the fuzzy sub module are high. Therefore， a neural fuzzy system combining Z-score and optimization technology is proposed to improve the calculation method of fuzzy rules. According to different methods to adjust the premise parameters and conclusion parameters， three hybrid algorithms are proposed： ZONFS1 based on BP + LSE， ZONFS2 based on LSE and ZONFS3 based on CGD + LSE. The experimental results show that compared with ANFIS and ZONFS1 algorithms， the time consumption of ZONFS3 algorithm is reduced by 37%， and the accuracy is improved by 26% compared with DTR and other algorithms； Compared with ZONFS3 algorithm， ZONFS2 has less time-consuming but lower accuracy； The average total score of ZONFSi algorithm is about 10 points higher than ANFIS； Compared with ANFIS algorithm， ZONFSi algorithm has higher accuracy and less time consumption， and has significant advantages in building depth model and processing high-dimensional data.关键词：neural fuzzy system;parameter optimization;fuzzy rules;Z-score;CGD
- 40
- |
- 11
- |
- 0
发布时间：2022-06-24 - 摘要：In order to solve the problem of false detection between some similar vehicles and the influence of partial occlusion conditions on the detection accuracy， an improved Faster R-CNN vehicle recognition algorithm is proposed. In this method， reasoning RCNN is introduced to give the network reasoning ability by combining with knowledge map， and on this basis， hole convolution is added to improve the receptive field， and the hole space pyramid is pooled to enhance the extraction of multi-scale information. The experimental results show that the improved Faster R-CNN model has high sensitivity for vehicle type recognition in complex scenes， and its map value reaches 93.86%， which has certain practicability.关键词：deep learning;object detection;convolution neural network;knowledge graph
- 81
- |
- 20
- |
- 0
发布时间：2022-06-24 - 摘要：For Francis turbine， the pressure pulsation of draft tube is an important factor leading to unit vibration and affecting the stable operation of unit. In order to effectively predict the pressure pulsation of draft tube of Francis turbine in the unit design stage， and take relevant measures to reduce the pressure pulsation. The prediction method of BP neural network is introduced to predict the pressure pulsation， and then the weight and threshold of BP neural network are optimized by using the global search ability of thinking evolutionary algorithm （MEA）. The results showed that the prediction accuracy of BP neural network optimized by MEA can reach 0.991 48， which was 0.721% higher than that of traditional BP neural network. The pressure pulsation prediction effect of BP optimized based on MEA was better than that of traditional BP， and the prediction accuracy was higher， which can be used in practical engineering.关键词：draft tube;pressure pulsation;mind evolutionary algorithm （MEA）;BP neural network;pressure pulsation prediction
- 20
- |
- 5
- |
- 0
发布时间：2022-06-24 - 摘要：The movement of biological clusters is very common in nature， and studying the behavior of biological clusters is the key for human beings to control complex systems of clusters. In view of real fish swarm motion data， we design an attention model suitable for fish swarm motion analysis by referring to the attention mechanism. By effectively encoding the observed information of self and neighbors， the attention network model is constructed to obtain the appropriate attention weight， and the coded information of neighbors is weighted and decoded by the attention weight， and the single motion decision is output. The macroscopic motion characteristics of the model were compared with those of real fish， and verified by multi-agent simulation experiments. The experimental results show that the proposed method is consistent with the experimental results of real fish， and can effectively drive agents to simulate fish swarm behavior， which lays a foundation for the large-scale realization of multi-agent swarm motion.关键词：complex system;cluster movement;deep learning;attentional mechanism
- 19
- |
- 7
- |
- 0
发布时间：2022-06-24 - 摘要：Aiming at the low accuracy of single convolutional neural network in recognizing gesture images in multiple complex backgrounds， a gesture image recognition method based on improved Xception network was proposed. Dense connections are used to replace residual connections， so as to retain the effect of jumping connections and reduce the number of depthwise separable convolution modules and network channels， so as to ensure effective utilization of network parameters and reduce the size of the model. SE module is fused to strengthen important features， and Feature Pyramid Structure is fused to obtain feature tensor containing multi-scale semantics， which is helpful for network classification. The results show that the number of parameters in the improved network is 1/5 of that in the original Xception network， and the accuracy is 99.64% in the NUS-Ⅱ（National University of Singapore） gesture dataset， which is 1.09% higher than that in the Xception network. The accuracy of Sign Language for Numbers gesture dataset is 99.7%， which is 0.15% higher than Xception network. Compared with ResNet50， DenseNet121， InceptionV3 and other commonly used gesture recognition networks， the recognition performance of the improved network was verified to be better， considering four factors including model training time， model size， number of computing parameters and recognition accuracy. The proposed gesture recognition method based on improved Xception network has high gesture recognition accuracy and strong generalization under the interference of multiple complex background factors. Meanwhile， the number of model parameters is less， and the comprehensive performance is better than many commom networks.关键词：Xception network;dense depthwise separable convolution module;SE module;feature pyramid structure;gesture recognition
- 33
- |
- 7
- |
- 0
发布时间：2022-06-24 - 摘要：With the rapid development of science and technology， people's demand for home furnishing is constantly increasing. With the penetration of deep learning， the development of smart home ushered in a new wave.The existing smart home system mostly carries out the corresponding operation according to the specific instructions of the user， and lacks the analysis of the emotion implied by the user's semantics， nor can it predict the user's behavior.Therefore， algorithms in the field of natural language processing in deep learning are used to establish semantic sentiment analysis model， and LSTM-RNN network is used to learn semantic text feature vectors after word segmentation， and attention mechanism is introduced to assign different weight values to semantic word vectors， so as to enhance feature training.The living room system can be accompanied by the user's voice instructions and mining the user's emotional information to make adaptive changes， enabling users to immerse themselves and experience more intelligent services.关键词：smart home;speech recognition;attention mechanism;LSTM-RNN;sentiment analysis
- 15
- |
- 5
- |
- 0
发布时间：2022-06-24 - 摘要：In order to solve the problems of poor robustness and sensitive preset parameters due to the fixed size of anchor frame in the existing vehicle re recognition based on anchor frame， a vehicle re-identification model without anchor frame based on improved fully convolutional one-stage object detection is proposed. Based on the classical network model， a multi-level feature module for aggregating feature pyramid network is proposed， and the output of the last layer of the feature pyramid network is used as the final re-identification feature. The validity of the model is verified on the VeRi-776 dataset. The experimental results show that vehicle re-recognition model based on improved FCOS has excellent performance in average precision， Rank-1 and Rank-5， and the recognition accuracy reaches 82.6%， 96.8% and 98.3%， respectively. Compared with the classical re-identification methods， the improved FCOS method is obviously better than the method based on two-stage detector in search accuracy.关键词：vehicle re-identification;fully convolutional one-stage object detection;feature pyramid network;convolutional neural networks
- 13
- |
- 6
- |
- 0
发布时间：2022-06-24 -
Research on the Agricultural Question Classification Technology Based on Deep Pre-training Model AI导读
摘要：As one of the core modules of agricultural intelligent question answering system， agricultural question classification has a decisive impact on the retrieval efficiency of question answering system. Aiming at the problems of sparse and nonstandard features of agricultural questions， construct a classification model of agricultural questions based on deep pre training methods （BERT， ERNIE）. Three groups of experiments were designed under the five categories of crops， horticulture， breeding technology， fisheries and agricultural engineering， and 125 000 agricultural data sets were used for experimental verification. And explore and typical deep learning methods （TextRNN Attention， Transformer） for analysis and comparison， and explore the impact of increasing data sets on classification results. The experimental results show that the classification effect of agricultural questions based on deep pre training language model is the best， and the F1 value of Ernie model in the test set is up to 94.76%， which is 1%~3% higher than that of other models. In addition， with the increase of data sets， the F1 values of each model are improved. ERNIE model can automatically classify farmers' questions， and has better text classification effect of agricultural questions.关键词：agricultural question classification;text classification;pre-training language model- 36
- |
- 9
- |
- 0
发布时间：2022-06-24 - 摘要：For the current situation that the supervision effect is affected by the supervision behavior of site personnel， psychological factors are considered in the model， prospect theory is introduced， and value functions are used to analyze the optimal supervision behavior of construction site safety officers and workers under different conditions. The influence of reward and punishment amount and reward and punishment strength on the evolutionary trend is explored through computer simulation. The results showed that： penalty amount P=10 and reward amount Q=10， penalty intensity
- 16
- |
- 6
- |
- 0
发布时间：2022-06-24 - 摘要：In the multi-AGV（Automated Guided Vehicle） logistics scheduling operation of intelligent factory， there are high requirements for scheduling optimization， which needs to consider the problems such as feeding type， high frequency， strong interference and complex moving line. Therefore，study the simulation of multi-AGV logistics scheduling in intelligent factory from the perspective of man-machine cooperation. Firstly， the simulation requirements are analyzed and defined， and the simulation content is determined based on the combination of spatial location setting and simulation logic setting. Finally， with an intelligent factory as the research object， the logistics scheduling of the factory is simulated and analyzed， and the operation process of AGV and the loading process of trailer and cage car are simulated， to optimize the logistics scheduling. The simulation study shows that， when different numbers of AGVs are deployed in the same area， the downtime and other material duration of the production line are significantly different. The deployment of different numbers of trailers will affect the material renewal duration of the AGV line warehouse， and the supporting number of forklifts in each berth will significantly affect the truck parking time of the corresponding berth. The simulation research can provide decision-making data and solution support for multi-AGV logistics scheduling in intelligent factory.关键词：system simulation;logistics scheduling;man-machine cooperation;multi-AGV
- 27
- |
- 9
- |
- 0
发布时间：2022-06-24 - 摘要：In order to verify the adaptability of shipborne magnetic resonance equipment in the vibration environment of ships at sea， numerical calculation and finite element simulation analysis of shipborne magnetic resonance equipment were carried out. First carry out finite element modeling and calculation mesh division of the hull system， use ANSYS Workbench to extract the mode of the overall structure， calculate the vibration frequency response transfer function under the action of the exciting force generated by the propeller and main engine and other excitation sources；Then the vibration response is numerically calculated and simulated， and the vibration isolation optimization device is designed according to the calculation results. The experimental results show that the vibration response of the shipborne magnetic resonance equipment installation position below 21Hz shows an overall upward trend， and the vibration response above 21Hz maintains a high energy vibration level. The vibration response curve of the four corners reaches a peak at 21Hz， and the vibration response is stable in the range above 30Hz. Therefore， the vibration input of the shipborne magnetic resonance equipment at the installation position of the ship is within a safe range， which can ensure the safety and reliability of the equipped ship's vibration.关键词：constant conductance ultra-low field;shipborne magnetic resonance equipment;ship vibration response;finite element numerical calculation;ship vibration isolation optimization
- 46
- |
- 13
- |
- 0
发布时间：2022-06-24 - 摘要：In order to effectively improve the efficiency of particle swarm optimization （PSO） algorithm in searching the optimal solution， an adaptive simulated annealing particle swarm optimization （ASAPSO） algorithm based on Metropolis criterion is proposed. Firstly， ASAPSO adopts a new adaptive extreme inertia weight method， which can effectively balance the global search and local search process of particle population. Secondly， by analyzing the information exchange mode between particle individuals， the central particle of particle flight learning exchange is constructed， which can effectively enhance the social learning ability of particle individuals. Finally， based on the simulated annealing selection probability of Metropolis criterion， the particle group is approaching the global optimal solution under the guidance of the central particle， which can effectively avoid the particle group falling into the local optimal region. The simulation results show that compared with other test algorithms， ASAPSO has obvious advantages in high convergence accuracy. In a variety of standard test functions， the convergence rate （CR） of ASAPSO can reach more than 94%， which can effectively improve the optimization efficiency of particle population.关键词：particle swarm optimization algorithm;simulated annealing;Metropolis criterion;ASAPSO algorithm
- 10
- |
- 4
- |
- 0
发布时间：2022-06-24 - 摘要：In order to improve the defect of traditional function point method in software evaluation at the early stage of the project， put forward a method based on data mining， which can quickly realize the software workload evaluation in the early stage of the project. The specific method is to identify the function features and analyze the workload of the function points of the historical software project in a specific field， and establish the workload weight relationship of different function points， so as to help the quick workload assessment of the new project.A total of 2000 function point data of several projects in a specific software field are summarized for analysis and evaluation. Finally， 1 649 standard function point workload data are cleaned out， and the workload evaluation model is established. Then a new project to be evaluated is used to verify the accuracy of the model. The experimental results show that the accuracy of the method is as high as 74.7%. With more and more accumulated historical project data， the workload weights of function points with different functional characteristics will be updated iteratively with higher accuracy， so as to improve the accuracy of workload evaluation in the early stage of the project.关键词：data mining;function feature recognition;workload weight;workload evaluation
- 16
- |
- 5
- |
- 0
发布时间：2022-06-24 - 摘要：The filtered-S least mean square （FSLMS） algorithm can be used to solve the nonlinear problem in active noise reduction. The MCNFSLMS algorithm combines two FSLMS filtering structures through joint parameters， achieving fast convergence and low steady-state error.However， the performance of the algorithm is affected by the lack of cross terms and the symbol judgment of the joint parameters in the iteration process. Therefore， a convex combined algorithm based on FSLMS and generalized functional link artificial neural network （GFLANN） is proposed. The cross term is introduced through GFLANN structure， and a new function is used as the joint parameter to avoid the judgment of symbol. Under three different nonlinear conditions， it is verified that the proposed combined structure has better performance than a single structure， and compared with the combined structure MCNFSLMS， the algorithm has improved noise reduction performance by 2dB， 1dB， and 4dB respectively ， at the mean time， reducing the amount of calculation， and having a lower steady-state error.The experimental results verified that this algorithm can better solve the nonlinear problem in active noise reduction.关键词：nonlinear active noise control;FSLMS filter structure;GFLANN filter structure;convex combination algorithm
- 12
- |
- 5
- |
- 0
发布时间：2022-06-24 - 摘要：With the wide use of Web services and the rapid development of SOA， the continuous improvement of resource sharing leads to a large number of different types of functional components. In order to quickly find the functional components that meet the requirements， realize dynamic binding， improve reusability and quickly rebuild the system when the requirements change， this paper proposes a dynamic binding framework of functional components based on semantic reasoning by studying the multiple invocation methods of Web services， unified invocation of functional components based on WSIF， semantic network and OWL-S， and realizes the unified description of functional components， Unified call.The framework can dynamically discover and bind the functional components that meet the requirements， realize the rapid construction of application system according to the requirements， and avoid the steps of manual comparison of service description and rebinding functional components， so that the construction efficiency of application system is 4 times higher than that of traditional methods.关键词：functional components;WSIF;semantic network;dynamic binding
- 15
- |
- 5
- |
- 0
发布时间：2022-06-24 - 摘要：Aiming at the anti salient characteristics of interior permanent magnet synchronous motor （IPMSM） and the defects of traditional genetic algorithm parameter identification method， a parameter identification method based on local search-based hybrid genetic algorithm （Is-hGA） is proposed. The Is-hGA is a combination of hill-climbing method and genetic algorithm. This method can identify four parameters of stator resistance， d-axis inductance， q-axis inductance and permanent magnet flux linkage at the same time. Generally， the parameter identification problem of IPMSM can be transformed into a problem to find the optimal solution. In the proposed hybrid optimization method， the global search task is performed by GA， while the hill-climbing algorithm is used for local search. In this way， the problem of poor local search ability of GA can be improved， but also the computing time can be saved greatly. The experimental results show that the minimum identification accuracy of the IPMSM parameter identification using the Is-hGA algorithm reaches 6.82%， 3.62%， 2.32%， and 2.12%， respectively. Compared with the traditional genetic algorithm， the search efficiency and identification accuracy are greatly improved. Therefore， the parameter recognition based on the Is-hGA algorithm can not only reduce the number of iterations， save computational time， but also have a high parameter recognition accuracy.关键词：interior permanent magnet synchronous motor;parameter identification;genetic algorithm;hill-climbing method
- 9
- |
- 3
- |
- 0
发布时间：2022-06-24 - 摘要：With the rapid development of economy， people's living standards have been greatly improved， and began to pursue a more ideal community living environment. At present， the construction of smart community is being carried out fervently. The monitoring and analysis of the community environment will help guide the community managers to make correct decisions on the environmental conditions， so as to improve the living conditions of residents. The research starts with the complex relationship of multiple environmental factors in the community， uses sensors to collect key environmental parameters such as temperature and humidity， light intensity， PM2.5 and noise in the community in real time， combines multi-data fusion technology， and adopts improved D-S evidence theory to make global fusion decision-making， realizing accurate decision-making for intelligent control of community environment. Compared with the decision-making model using single-sensor data， multi-data fusion technology effectively combines the different effects of various environmental factors and improves the accuracy of decision-making. Compared with the results of traditional DS algorithms， the support for correct decisions has increased by 39%.关键词：smart community;environmental monitoring;multi-data fusion;intelligent control;D-S evidence theory algorithm
- 16
- |
- 4
- |
- 0
发布时间：2022-06-24 - 摘要：Shield tail is a thin-walled component， it is easy to deform under soil pressure. In order to monitor the deformation of shield tail in real time， a remote monitoring method of shield tail deformation is proposed， and the corresponding monitoring system is developed. The system mainly includes strain acquisition module， cloud computing center and visualization module. Through the strain acquisition module， the collected original strain data is transmitted to the cloud computing center through TCP protocol for signal filtering， and the concurrent multi-threaded server is used to improve the data throughput. The calculated deformation data is visually displayed in the form of Web pages. After three months of practice， it is proved that the Web terminal still maintains a low delay under high load， the average response time is 150ms， and the system is stable and reliable.关键词：shield tail of shield machine;telecommunication;data processing;Python visualization
- 34
- |
- 9
- |
- 0
发布时间：2022-06-24 - 摘要：In order to solve the increasingly prominent problems of home-based elderly care and care for the elderly population， a wearable system based on acceleration sensor and height sensor is designed. By fusing multi-sensor parameters with Kalman filter and combining quaternion to recognize human posture， fall detection is realized and abnormal behavior is monitored in real time. The experimental results show that the Kalman filter based on multi-sensor parameter fusion and the analytical algorithm of Mahony attitude angle can detect falls in the test samples with an accuracy of 95%. It has the characteristics of high detection accuracy， small amount of calculation and convenient detection， and can better solve the problems of home care and nursing.关键词：wearable;multi-sensors;Kalman fusion;quaternion;attitude angles
- 13
- |
- 7
- |
- 0
发布时间：2022-06-24 - 摘要：In order to overcome the disadvantage that the subdivision of the existing detection methods consumes a lot of FPGA hardware resources and is not conducive to the subsequent galvanometer control software development. Based on a 13 bit sin / D Converter IC with signal calibration_ NQC combines Mercury_1200 sine cosine relative encoder， a digital galvanometer position detection platform is built. BISS is written by programmable logic device FPGA and hardware description language_ C serial communication and decoding module， by BISS_ The C serial communication module transmits configuration data to the IC_ NQC reset and register configuration. IC_ NQC performs digital quantity processing on the sine cosine coded signal to extract the position information. BISS_ C serial communication module acquisition IC_ The position data information of NQC is decoded to obtain the galvanometer position. The experimental results show that this method greatly reduces the consumption of FPGA hardware resources， the position acquisition cycle is 28.1us， can realize 1 024 subdivision and meet the requirements of high-precision digital galvanometer control for position acquisition.关键词：IC_ NQC;digital galvanometer position detection;BISS_ C serial communication;FPGA
- 11
- |
- 4
- |
- 0
发布时间：2022-06-24 - 摘要：The fault diagnosis method of motor bearing based on data-driven often needs a large number of fault samples to obtain the ideal diagnosis effect. However， in actual industrial production process， the fault samples are often difficult to obtain. In order to solve the problem of limited number of fault samples， propose a data augment method which integrates the squeeze and exception （SE） mechanism into auxiliary classifier generative adversarial nets （ACGAN）， so as to further improve the consistency of the generated samples and the original samples. The simulation results show that the proposed method can further improve the accuracy of fault diagnosis compared with the traditional model， and the highest accuracy can reach 99.82%. The improved ACGAN can learn the features of the original samples more effectively， which further improves the accuracy of fault diagnosis. It has the advantages of fast convergence speed and more stable quality of generated samples.关键词：data-driven;fault diagnosis;ACGAN;data augment;SE
- 16
- |
- 5
- |
- 0
发布时间：2022-06-24 - 摘要：To improve the analysis ability of public opinion information， we design a mean shift algorithm based on the Spark framework. For public opinion， using the Ansj word segmentation and Word2vec algorithm feature extraction， finally clustering based on the Spark framework parallel computing model and the principle of mean shift algorithm. The numerical results show that， in both Iris and Wine data sets， the accuracy of the mean shift algorithm is over 90%， the clustering result is significantly better than the K-means algorithm， then the mean shift algorithm has better adaptability. In the performance experiment， it can effectively improve the operation efficiency of the algorithm and has better data scalability by increasing the degree of parallelization of the algorithm operation program. Therefore， the algorithm can effectively improve the analysis ability of public opinion，and help establish a healthy network environment.关键词：public opinion;Spark;mean shift;clustering;parallelization
- 25
- |
- 4
- |
- 0
发布时间：2022-06-24 - 摘要：Through text mining on the collected customer online comments， the influencing factors of customer online shopping satisfaction are extracted to provide decision-making for enterprise product development and online marketing.On the basis of grasping the online reviews of JD e-commerce customers by web crawler technology， LDA model is used to extract topics from the cleaned online reviews data， and then cluster analysis is used to obtain the influencing factors of customer online shopping satisfaction. The results show that the product quality， product price， product design， logistics service and brand reputation are important factors affecting customer online shopping satisfaction.关键词：online shopping;customer satisfaction;LDA model;text mining
- 28
- |
- 5
- |
- 0
发布时间：2022-06-24 - 摘要：In view of the difficulties of individuals and small enterprises in bank lending， as well as the greater capital risks borne by banks， the staff are more subjectively affected in lending audit. Using blockchain technology， this paper studies and proposes a bank financial lending scheme based on smart contract. Firstly， EBBI （Ethereum-Borrower-Bank-Investor） lending and investment model and intelligent risk control evaluation algorithm based on deep learning are designed in this scheme； Secondly， intelligent contract algorithms such as new subject matter， investment， repayment and compensation are designed and verified；Finally， a case study of small and micro enterprise loans is carried out.The experiment uses the public data set of paipai loan， and the accuracy of the algorithm is 94.33%； In 500 transactions， the average query processing time of the system is 0.23s. The experimental results show that the intelligent risk control evaluation model in this scheme has reached a high reliability level； The implementation of smart contract algorithm meets the requirements of system operation efficiency and can help the borrower borrow quickly without affecting the original credit business of the bank.关键词：ethereum;blockchain;smart contract;bank loan;intelligent risk control
- 36
- |
- 8
- |
- 0
发布时间：2022-06-24 - 摘要：The biggest drawback of the IoT is the lack of mutual trust between devices. In order to ensure the security of data and devices within the IoT system， all IoT devices in the system need to be authenticated first. Once the central database fails， the entire IoT will be paralysed. Based on the requirement analysis of data security storage system for IoT devices， a blockchain-based IoT smart contract model （BC-SC model） is designed to establish a trustworthy transaction environment using blockchain peer-to-peer distributed architecture to realise machine-to-machine （M2M） and human-to-machine mutual trust models. Among them， the BC-SC model distinguishes the nodes in the network according to their storage capacity as primary and secondary nodes， and the primary node realises the data verification function in the system， and distributes keys with identity identification to the light nodes in combination with the State Secrets SM9 algorithm to achieve M2M two-way authentication； in order to ensure the trustworthiness of the data， it is ensured by choosing a hybrid encryption method in this model， and at the same time， according to the flexible and secure smart contract In order to ensure the trustworthiness of the data， it is ensured by choosing a hybrid encryption method in this model， while the deployment of IoT devices is completed according to the flexible and secure characteristics of smart contracts， thus enhancing the security of the whole system.关键词：blockchain;IOT;smart contract;SM9;M2M
- 31
- |
- 9
- |
- 0
发布时间：2022-06-24 - 摘要：Container technology is the foundation to support modern cloud-native applications. Container cluster service is the mainstream model of container provision，and using container cluster services in a multi-cloud environment has become a common practice in enterprises. The key to constructing this type of service is the adaptation of multi-cloud environments， which specifically involves solving problems such as the heterogeneity of the host environments， the network access constraints between multiple clusters， and the diversity of container loads. To solve the above problems， OMCC cloud container service system was built.The system is a practice of building a container system for multi-cloud. It effectively solves several adaptability problems of container cluster services in a multi-cloud environment through mechanisms such as the bootstrapping container cluster creation， the proxy gateway and intranet penetration， and the unified load abstraction. The application experience of OMCC provides a general guideline for the research and development of container cluster services in the multi-cloud scenario.关键词：multi-cloud environment;container cluster;cloud service;platform software
- 150
- |
- 21
- |
- 0
发布时间：2022-06-24 - 摘要：The existing methods of regulating traffic are primarily based on unidirectional neural network prediction allocation， and the lack of edge information leads to low accuracy，therefore propose an intelligent traffic prediction method of communication base station based on bi-directional gate recurrent unit to solve this problem. This method selects gate recurrent unit to capture the potential law of the time series effectively， break through the shortcomings of the unidirectional allocation method， and train independently from the front and rear directions， to achieve a more accurate prediction effect with complete historical information. According to the 8GB cell uplink traffic data collected by Beijing guoce satellite mapping， the experiments are compared with long short-term memory， bi-directional long short-term memory and gate recurrent unit. The results show that the bi-directional gate recurrent unit proposed in this method has an average increase of about 0.1 in the R2 and significantly improved the prediction accuracy compared with the comparison algorithm. It plays a certain role in decision support for the regulation of base station traffic and has a particular practical significance.关键词：deep learning;artificial intelligence;base station traffic prediction;bi-directional gate recurrent unit;recurrent neural network
- 18
- |
- 4
- |
- 0
发布时间：2022-06-24 - 摘要：Nowadays， people's life is inseparable from online shopping. However， the traditional online shopping system still has some problems， such as low single machine capacity， unable to cope with high concurrency and insufficient cache. Therefore， improving the capacity of online shopping system， anti concurrency ability and strengthening cache can effectively improve the efficiency of query data. Therefore， optimize and reconstruct the online shopping system under the traditional stand-alone mode， designs a high availability framework based on distributed cluster， uses nginx access server to realize the high availability of the cluster and improve the anti concurrency ability， reduces the access pressure of customers to the server and improves the data reading efficiency by calling redis cluster cache through nginx Lua. Finally， the pressure measurement experiment is carried out. By observing the average response time and maximum throughput parameters， it shows that compared with the traditional system， the average response time of the optimized system under 1 000， 1 500 and 2 000 concurrency is reduced by 72ms， 151ms and 135ms respectively， and the throughput is improved by 624 / s， 1 274 / s and 1 062 / s. It is further verified that the optimized system has obvious improvement in anti concurrency and data extraction performance.关键词：distributed deployment;load balancing;Redis cluster;Nginx Lua cache;Performance stress test
- 24
- |
- 9
- |
- 0
发布时间：2022-06-24 - 摘要：Aiming at the problems of low contrast and large background noise of sediment adhesion in underwater dam crack image， combined with the different characteristics of fine linear crack and wide linear crack， an underwater crack detection method based on visual processing is proposed. The threshold segmentation algorithm combined with Prewitt operator edge detection is used to segment the small cracks in the underwater image. The morphological processing is used to eliminate the interference noise of the segmented underwater dam crack binary image and extract the target crack information. The adaptive threshold segmentation method combined with Gamma correction is used for multiple processing to obtain the best segmentation threshold of underwater dam crack image. The morphological processing method is used to extract the connected domain and fill the cavity of the crack binary image to obtain the complete crack binary image. Finally， according to the camera parameters of the underwater dam crack image， the physical size of each pixel is estimated， and then the characteristic parameter value of the dam crack is obtained. Compared with the existing crack detection algorithms， the proposed detection algorithm can accurately and effectively extract the small and wide linear crack information of the complete underwater dam in the complex underwater environment， estimate the length， width， area， direction and other parameters of the underwater dam crack， and has good robustness.关键词：underwater image;image segmentation;crack detection;adaptive threshold segmentation;morphological processing
- 19
- |
- 8
- |
- 0
发布时间：2022-06-24 - 摘要：Statistical methods were used to screen and analyze the brain network feature set and morphological feature set， and traditional machine learning methods were used to optimize， so as to transform multi-sequence features into specific indicators and construct sensitive and specific feature sets of brain aging.MethodsThe experimental data included brain imaging data of 96 normal subjects， which were divided into young group （48 cases） and old group （48 cases）. First， feature screening was carried out by one-way ANOVA， and then the classification experiment of young and old subjects was carried out based on machine learning classifier. Finally， the Receiver Operating Characteristic （ROC） and Area Under the Curve （AUC） values and Accuracy （ACC）， Specificity （SPE） of different models in corresponding feature sets are evaluated by the five-fold cross-validation method.ResultsThe highest classification accuracy was 70% and 74.4% for network feature set and morphological feature set respectively， and 90% for joint feature set. The results showed that the optimized feature subset extracted based on this method has high sensitivity and specificity to brain aging.关键词：magnetic resonance imaging;machine learning;brain aging
- 19
- |
- 3
- |
- 0
发布时间：2022-06-24 - 摘要：Scene text recognition is a very challenging task in recent years， aiming at the difficulty of recognition due to the variability and curvature of dense text in natural scenes. A correction and recognition method for scene image text is proposed. First， use the correction network to correct the scene text， and then input the corrected text into the dual-branch network module to extract image features for fusion. The dual-branch module uses InceptionV2 and variable convolution instead of ordinary convolution to obtain different ranges of receptive fields. Increase and adjust the direction vector of the convolution kernel to make the shape of the convolution kernel closer to the shape of the text. Finally， the textual semantic information with different weights is obtained through the two-way gated loop unit embedded in the attention. Experimental results show that the model has achieved convincing results on the ICDAR2013， ICDAR2015， and CUTE80 datasets， especially on the CUTE dataset of curved texts， with an accuracy rate of up to 89.54%， which is nearly improved compared to traditional methods. 1.9%， the model can effectively identify distorted text information.关键词：image processing;correction network;text recognition;natural scene
- 10
- |
- 3
- |
- 0
发布时间：2022-06-24 - 摘要：Traditional license plate recognition technology not only requires high hardware， but also has shortcomings in recognition accuracy and speed. In recent years， convolutional neural networks has attracted more and more attention in the field of image recognition. By improving and reconstructing the structure of AlexNet convolution neural network， a new license plate recognition method is proposed. The experimental results show that the method can avoid the impact of license plate character segmentation and complex background environment on the recognition results for the license plate recognition in natural scenes， and the recognition accuracy rate is as high as 98.03%， which has high reliability.关键词：AlexNet;license plate recognition;deep learning;convolutional neural network
- 26
- |
- 4
- |
- 0
发布时间：2022-06-24 - 摘要：Present a separable dual information hiding scheme in encrypted color image， so as to solve the difficult problem of copyright protection.Separate the color carrier image into RGB to obtain three channels of red， green and blue. In the blue channel， firstly， adopt all-bit plane rotation encryption to obtain the ciphertext domain， then use QR decomposition and singular value decomposition （SVD） to determine the hidden position， and then perform fruit fly optimization algorithm （FOA） to get hidden strength， finally， hide the copyright information which is scrambled by the proposed fibonacci-arnold scrambling （FAS）. In the red channel， firstly， the newly proposed two-layer visual cryptography method is adopted， then HOG feature extraction， boost normed singular value decomposition （BN-SVD） and FOA are used to extract feature vectors， and then LT coding is conducted to generate initial secret zero-information， finally， combine feature vectors， initial secret zero-information and timestamps to generate final copyright information. In this paper， even in the case of the maximum capacity， the average PSNR and SSIM between watermarked image and original image are 48.438 5dB and 0.959 5， respectively. It indicates that the proposed scheme holds better visual quality. Furthermore， it holds stronger robustness， and the average NC under various attacks is about 0.9.The proposed scheme realizes dual copyright protection. It achieves strong robustness and can resist geometric attack and noise attack. Therefore， the experimental data show that the proposed method not only has better invisibility and stronger robustness， but also has higher security， and better real-time performance， which is suitable for the copyright protection of high-security color digital images.关键词：image copyright protection;dual information hiding;ciphertext domain;separable
- 13
- |
- 10
- |
- 0
发布时间：2022-06-24 - 摘要：Computational thinking is an indispensable scientific way of thinking in the information society. It has become an important goal of college teaching to cultivate students' computational thinking ability. Based on the connotation of computational thinking and the advantages of PBL （Project-based Learning） teaching， take the pattern recognition course as an example to construct a PBL teaching mode oriented to the cultivation of computational thinking. This teaching method takes students as the main body， teachers as the leading， cultivating computational thinking as the goal， and projects as the mainline to carry out innovative and exploratory teaching and learning activities， aiming at improving students' computational thinking ability and hands-on practice ability， so as to promote students to achieve true deep learning.关键词：computational thinking;PBL teaching mode;pattern recognition
- 23
- |
- 13
- |
- 0
发布时间：2022-06-24 -
Cluster， Attribution and Countermeasure Research on Online Learning Behavior of College Students AI导读
摘要：Under the background of the comprehensive development of online education， taking the online learning behavior of college students as the research object， an investigation scale including five levels of learning motivation， external support， external supervision， learning behavior and learning effect was designed， and clusters， attributes and models the data results by using the methods of cluster analysis， analysis of variance and structural process model， From the two dimensions of resource interaction behavior and interpersonal interaction behavior， learners can be clustered into bystander learners， passive learners， active learners and interactive learners. Combined with teaching practice， relevant suggestions to improve the effect of online teaching are put forward.关键词：e-learning;learning characteristics;learning effect;learning analysis;causal attribution- 21
- |
- 8
- |
- 0
发布时间：2022-06-24 - 摘要：In order to understand the research status and trend of artificial intelligence education in China， and provide suggestions for its research and development. Taking the journal papers of CNKI database from 2016 to 2020 as the data source， using Bicomb2 0、Ucinet6. 0 and SPSS26.0 software， cluster the topics of domestic AI education research， grasp the research status and development trend of AI education through multi-dimensional scale analysis， combined with the visual map of social network and keyword centrality analysis， and deeply analyze the research hotspots and the relationship between keywords. The results show that the research in the field of AI education in China mainly focuses on four aspects： theoretical concept， technical service， educational application and development direction， and puts forward suggestions on the development trend of talent training， curriculum， teacher team construction and ethics education in the future.关键词：artificial intelligence education;cluster analysis;multidimensional scaling analysis;social network analysis
- 17
- |
- 8
- |
- 0
发布时间：2022-06-24 - 摘要：The classic flipped class model is not effective enough in the teaching practice of the computer organization principle which involves many difficult and strong logic knowledge points. students' self-learning has certain blindness and low efficiency. Therefore， a problem-oriented FCM improved teaching model oriented to the computer organization principle is proposed， which is problem-oriented and based on the appropriate "online + offline" teaching design. Also， its learning process is with the clues of finding problems， analyzing problems， and solving problems. From the two dimensions of teaching model and assessment model， the proposed teaching model explains how to achieve the teaching goals， and provides teaching design and management references for the courses with abstract theories， strong logic， and difficult knowledge points.关键词：computer organization principle;self-learning;problem-oriented;FCM improved teaching mode
- 22
- |
- 16
- |
- 0
发布时间：2022-06-24 - 摘要：In view of the phenomenon that the safety talents of industrial control system are extremely scarce and the psychological problems of students occur frequently， by analyzing the teaching practice of the safety experiment course of industrial control system under the background of "new infrastructure construction"， constructs the safety experiment course of industrial control system and carries out teaching reform， so as to meet the social demand for the safety talents of industrial control system. First of all， clarify the talent training objectives， take the professional frontier as the core， and build a modular and matrix curriculum content system to avoid causing students' acquisition helplessness. Secondly， in terms of teaching methods， pay attention to the cultivation of students' happiness and resilience， guide students to master basic professional knowledge and have transferable skills and abilities， and explore diversified and accurate experimental teaching methods. Finally， the process assessment based on student value-added is designed and applied. Through the teaching of the whole life cycle training mode， promote the improvement of the safety teaching quality of industrial control system and cultivate more high-quality talents for the construction of "new infrastructure".关键词：new infrastructure;training structure;happiness psychology and resilience;transferable skills
- 12
- |
- 8
- |
- 0
发布时间：2022-06-24
- Technical support is provided by Beijing Founder electronics co., LTD 鄂ICP备12016542号-70
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0