最新刊期
卷 22 ， 期 3 ， 2023
- 摘要：In the digital environment， the multi-source heterogeneous mass data brings new challenges to the standardization of investment audit data， which further increases the difficulty of data collection， cleaning， conversion and utilization， and seriously affects the quality of investment audit. The main development direction of investment audit data governance is how to promote the integrated development of investment audit practice and big data technology， strengthen the centralized management of data resources， build a unified and shared comprehensive audit data governance framework， and realize a visual and intelligent audit mode covering the whole process of data governance. Based on the existing pain points of investment audit data management， explore the application mode of AI and Robotic Process Automation technology in the whole life cycle of audit data management.It can provide new ideas for promoting the collection of audit data resources， improving the standardization of audit data， innovating the mechanism of audit data fusion analysis and sharing exchange， and realizing the automation of data governance process.关键词：data governance;investment audit;robotic process automation;artificial intelligence
- 117
- |
- 15
- |
- 0
发布时间：2023-04-18 - 摘要：Building energy consumption prediction plays an important role in building energy management， energy conservation and fault diagnosis. However， there are nonlinear and outlier points among building energy consumption data， which leads to the decrease of energy consumption prediction accuracy. To solve the above problems， the MRGALnet building energy consumption prediction model based on feature extraction， clustering and improved LGBM is proposed. Firstly， the subsets of features that have the greatest impact on building energy consumption are selected through MI+RFE secondary feature selection algorithm. Secondly， building data with similar energy consumption characteristics are grouped by Gaussian mixture clustering algorithm. Thirdly， energy consumption data for each cluster are predicted by LGBM. Furter more， an adaptive loss function is designed to improve the prediction performance of LGBM. Through comparative experimental analysis， it can be seen that MI+RFE feature selection algorithm can effectively remove redundant features， GMM can reasonably cluster the original data， and ALGBM model can adaptively determine the hyperparameters of the loss function according to the energy consumption data of different clustering， so as to improve the model prediction performance. The MRGALnet model combined with the above algorithms is optimal in terms of prediction accuracy and convergence speed. The MRGALnet model integrating the above algorithms can further improve the prediction accuracy and convergence speed.关键词：building energy consumption prediction;feature selection;clustering;light gradient boosting machine;adaptive loss function
- 77
- |
- 12
- |
- 0
发布时间：2023-04-18 - 摘要：The massive amount of data generated by the development of big data technology urgently requires a reliable data storage method. The existing proactive failure prediction methods have achieved better performance than traditional reactive methods， but some problems still exist to be solved in the disk failure prediction. Most current methods are offline and divide data into samples through a sliding window with subsampling or oversampling to solve the imbalance problem. However， the distribution of hard disk data changes over time， subsampling may lose some valuable features， and oversampling may lead to training overfitting. Meanwhile， these methods ignore the temporal dependency of disk data. To solve these problems， the proposed method casts a hard disk's entire operation and maintenance process in the storage system as a markov decision process， which is solved using the reinforcement learning method DQN. Based on reinforcement learning， the imbalance problem turns into the sparse reward problem， which is solved by reward shaping and exploration. The replay buffer can update the model online by the stored tuples. Experiments on the open-source dataset BackBlaze verify the effectiveness of the proposed method.关键词：hard disk failure;failure prediction;deep reinforcement learning;DQN;reward shaping
- 79
- |
- 20
- |
- 0
发布时间：2023-04-18 - 摘要：A city is composed of regions with different mixed function distributions （urban form）. The functional structure of a region may include school districts， residential areas and other components. Understanding urban form plays an important role in urban planning and governance. Aiming at the problem that previous studies did not deeply combine the static structure and dynamic characteristics of the region， a new quantitative method of urban form is proposed. With the idea of topic modeling， structured check-in patterns are regarded as regional documents； The similarity of static structures （including categories， consumption costs and service satisfaction of POIs） between regions is regarded as links between documents； The method used minimum spanning tree to divide a city into regions. The static similarity between regions is measured by an algorithm based on fuzzy C-means. And the relational topic model （RTM） is used to reveal the dynamic internal relationship between human social activities and urban form， and further combined with the static information of POIs. In the end， the mixed function distribution of each region of the city is obtained. Conducted on real data sets in New York for calibration， the experimental results show that， compared with the existing models， the experimental error of the proposed method is smaller and the effect is better.关键词：urban form;data mining;machine learning;points of interest;check-in data
- 47
- |
- 8
- |
- 0
发布时间：2023-04-18 - 摘要：Aiming at the problem of data redundancy and low efficiency in knowledge retrieval optimization of knowledge base， a knowledge retrieval optimization algorithm based on improved random forest was proposed to improve the optimization efficiency of knowledge retrieval. This algorithm adopts mutual information weighting method and combines with particle swarm optimization algorithm to determine the weight of decision tree according to the relevance of decision tree and the evaluation accuracy， so as to optimize the structure of forest， and finally obtain the algorithm model with lower coupling degree and higher evaluation .Experiments show that the NDCG index of knowledge retrieval quality of the algorithm is improved by 14.25% on average， and the mean index of average evaluation accuracy（MAP）is improved by 13.75%， indicating that the algorithm can improve the retrieval quality and solve the problem of low optimization efficiency and accuracy of retrieval results.关键词：random forest;knowledge retrieval;knowledge base;optimization algorithm;search sort;mutual information
- 23
- |
- 3
- |
- 0
发布时间：2023-04-18 - 摘要：Aiming at the problem that the existing algorithm has insufficient detection accuracy and slow detection speed for targets such as vehicle pedestrians in traffic scenarios， the backbone feature extraction network of YOLOv4 is replaced by MobileNetV3 structure in YOLOv4， and the amount of trunk network parameters is greatly reduced through deep separable convolution in MobileNetV3， and then the SPP structure is transformed by using the structure of the cross-stage local network to eliminate information redundancy. Finally， the attention module CBAM is added to the network structure to improve the detection accuracy. Through experiments on the UA-DETRAC dataset， the average detection accuracy of the method reached 92.6% with 24% reduction of model size， compared with YOLOv4， the average detection accuracy was increased by 1.7%， the average missed detection rate was reduced by 11%， and the recall rate was increased by 1.6%. The results show that the algorithm is feasible in optimizing traffic scene target detection.关键词：object detection;lightweight construction;attention mechanism;cross stage network;feature extraction
- 39
- |
- 13
- |
- 0
发布时间：2023-04-18 - 摘要：Core meaning of the new development pattern is to promote the modernization of the national governance system and governance capacity by taking the domestic big cycle as the main body and promoting the mutual promotion of the domestic and international double cycles. This requires strengthening the new infrastructure of data technology and digital development and application， and promoting the construction of digital society and digital government. In recent years， the number and scale of government big data application informatization projects have shown explosive growth， but generally belong to high-investment， long-cycle and high-risk projects. Whether the expected application results can be achieved has become a problem that governments at all levels must pay attention to. To this end， based on the analysis of the requirements for building a new development pattern proposed in the "Fourteenth Five-Year Plan" and the characteristics of governance in the digital era under the background of the "Post New Movement" theory， propose the necessity of performance audit of government big data application projects， analyze the uniqueness of big data application projects compared with general government information projects， and propose a new perspective of performance audit of government big data application projects， The construction of a two-dimensional supervision model based on "process indicators" provides a general idea for the performance audit of big data application projects.关键词：new development paradigm;post new movement;digital governance;big data application projects;performance audit;digital technology
- 11
- |
- 4
- |
- 0
发布时间：2023-04-18 - 摘要：Deep learning-based cyber attack detection methods have been rapidly developed with their powerful feature representation and extraction capabilities. However， deep learning-based cyber attack detection models built in traditional Euclidean space cannot effectively capture the rich topology in the attack graphs. In order to efficiently model the potential data patterns in the cyber attack graph and thus improve the detection accuracy， a cyber attack detection model based on dual curvatured Riemannian manifold is proposed. this model migrates the cyber attack detection model in the traditional Euclidean space to a heterogeneous non-Euclidean representation space， and uses the large-scale hierarchical and circular graph structure pattern modeling capabilities possessed by the non-Euclidean curvatured Riemannian geometry space to obtain high-quality attack graph representation vectors， thereby improves the cyber attack detection accuracy. The experimental results show that Precision，Recall，F1-score of the proposed dual curvatured Riemannian manifold-based cyber attack detection model were 0.963，0.964，0.964，respectively，which can effectively analyze and detect attacks.关键词：cyber security;attack detection;Riemannian manifold;curvatured space;neural network
- 18
- |
- 5
- |
- 0
发布时间：2023-04-18 - 摘要：At present， some progress has been made in detecting DDoS attacks at the application layer， but these methods can only detect one or several specific types of attacks and cannot adapt to the complex and changeable mixed attacks in the application layer. Therefore， a class of detection methods that adapt to hybrid attacks is urgently needed. Aiming at the deficiencies of DDoS attack detection methods in the current application layer， an application layer DDoS attack detection method based on the RF-SVM model is proposed. Firstly， the random forest feature importance algorithm evaluates the importance of specific features under mixed attacks. Then the feature dimension is reduced according to the FDRCA algorithm， and finally， the SVM classifier is used for classification. After experimental verification， the RF-SVM model can adapt to the application layer's complex and changeable mixed attacks. By comparing the BP algorithm， the single SVM algorithm， and the HsMM algorithm， it can be verified that the RF-SVM model has a higher detection rate and better mixing attack adaptability.关键词：distributed denial of service;support vector machine model;random forest;feature importance assessment
- 20
- |
- 7
- |
- 0
发布时间：2023-04-18 - 摘要：With the development of bionics related disciplines such as biology and optoelectronics， the research field of bionics is gradually expanding， and bionics researchers' demand for bionics knowledge prediction and other applications is also increasing. However， the lack of knowledge sharing and technology reuse among various fields of bionics has become an obstacle to the development of bionics. Therefore， it is necessary to deal with the requirements of knowledge management in the field of bionics and reduce the complexity of bionics research. After analyzing and investigating the characteristics of bionic data and knowledge graph technology， the construction framework of knowledge graph in bionic field is proposed according to the core construction steps of bionic field characteristics. The experiment shows that the established knowledge graph construction process model has excellent performance， which proves that the bionic knowledge graph can integrate， represent and manage the bionic technology， and can adapt to the task scenes in the bionic field.关键词：bionic;knowledge graph;entity;biology
- 27
- |
- 4
- |
- 0
发布时间：2023-04-18 -
Reversible Information Hiding Based on Secret Sharing of Multiple Hiding Devices in Medical images AI导读
摘要：Cloud storage and other technologies not only bring convenience， but also bring huge risks in terms of security. Reversible information hiding technology has attracted more and more attention from researchers and has been widely used to protect the privacy of users' cloud server data. Aiming at the problem that medical images are vulnerable to attack， a reversible medical image information hiding algorithm based on secret sharing and contrast enhancement of multiple hiders is proposed. Firstly， histogram stretching is used to enhance the contrast and improve the embedding capacity and image quality. Then， different histogram embedding methods are used to hide the private data in the ROI/RONI region of the medical image. Finally， the secret sharing technology is used to encrypt the image into a number of different encrypted images of the same size， and then send them to a number of different data hiders for data hiding. Even if some data hiders are attacked or betrayed， the original image can be recovered by collecting a sufficient number of marked encrypted images from the remaining data hiders. The experimental results show that the embedding rate of the proposed algorithm is improved by 568.56% at the highest and 33.39% at the lowest compared with the related algorithm. After encryption， Share image correlation is as low as 0.001 2； The average values of RCE， PSNR， SSIM and NC with dense images are as high as 0.524 5， 30.08， 0.949 1 and 0.999 6. Compared with related algorithms， this algorithm has obvious advantages in comprehensive performance.关键词：reversible data hiding;secret sharing;encrypted images;privacy protection- 12
- |
- 2
- |
- 0
发布时间：2023-04-18 - 摘要：Predicting the cloud computing resource can highly ehance the ability to deploy ahead of data center，but it has periodic and aperiodic change patterns. In order to accurately analyze the load changes of cloud resources and improve the performance of CPU load prediction in cloud resources， we propose a combined model based on decomposition-prediction. In the decomposition part， to achieve adaptive acquisition of relatively stable multiple components， we decompose the original time series based on the variational mode decomposition method （VMD）， and innovatively consider the minimized sample entropy as the objective function of the whale optimization algorithm （WOA）， which performs proxy search for VMD parameters， then we remove the interference components in the high-frequency and low-frequency components by calculating the variance contribution and similarity. In the prediction part， the features are extracted by the convolution layer， BiGRU is used to learn the bidirectional features and dependencies of the sequence. Finally， the attention mechanism is integrated into the weighting to obtain more accurate prediction results. Taking the CPU usage of machines in Alibaba Cloud Cluster Data as an example， compared with the GRU， CNN-BiLSTM， GRU-LSTM and WVMD-BCGRU， the RMSE of the proposed model is reduced by 20%， 14% ， 7% and 3%， respectively， which has higher prediction accuracy and more stable prediction performance.关键词：cloud resource load prediction;VMD;attention mechanism;GRU;WOA;SampEn
- 19
- |
- 3
- |
- 0
发布时间：2023-04-18 - 摘要：In order to solve the problem of low fault diagnosis accuracy of rolling bearings under variable working conditions such as different speeds and dynamic loads， a multi-channel time convolution network （MTCN） is proposed based on the existing time convolution network （TCN） to obtain the characteristics of vibration signals under different receptive fields， thereby improving the accuracy of rolling bearing fault diagnosis. First， the network uses three TCN networks， two of which are original vibration signals input to TCN networks with different expansion scales， and the other is to input the extracted time-domain features into TCN networks. Then， the three features are spliced and input into the full connection layer for multi classification through Softmax. The experiment shows that the accuracy of rolling bearing fault diagnosis of MTCN can reach 97.19% in the data set containing multiple rotating speeds， dynamic loads and other working conditions， which is higher than that of long short memory network （LSTM） and one-dimensional convolution AlexNet.关键词：deep learning;multipath time convolution network;dynamic load;rolling bearing;fault diagnosis
- 19
- |
- 5
- |
- 0
发布时间：2023-04-18 - 摘要：Aiming at the problem that no research has yet used deep learning methods to integrate gallbladder cancer imaging， radiomics features and tumor markers and other laboratory test data into the diagnosis of gallbladder cancer， an algorithm for gallbladder cancer diagnosis based on cross-modal feature fusion was proposed. Firstly， the 3D U-net network is used to segment the gallbladder region and extract the radiomic features， and the 3D convolutional neural network is used to extract the depth features of medical images. Then the two are combined with the laboratory examination data， and the fusion result is used as a classifier. input for gallbladder cancer diagnosis. The experimental results show that the method improves the classification accuracy， specificity， sensitivity and precision by about 16.67%， 12.62%， 11.54% and 13.14%， respectively， compared with the single-class feature model with the best classification. Compared with the five common image classification models， the accuracy， specificity and precision are improved by at least 10.00%， 25.00% and 13.33%. Therefore， it is concluded that this method has better accuracy and reliability in the diagnosis of gallbladder cancer.关键词：gallbladder cancer;cross-modal feature fusion;3D U-net;radiomics;3D convolutional neural network;laboratory inspection data
- 16
- |
- 2
- |
- 0
发布时间：2023-04-18 - 摘要：The construction of the knowledge graph of the diagnosis and treatment of lung cancer by famous and old Chinese medicine practitioners has significant advantages in correlation， expression， mining and utilization of knowledge， but the knowledge records of lung cancer cases are scattered， there are many tacit knowledge， and there is also rich relationship knowledge. In order to solve this problem，use the RED-GNN model to infer existing facts based on the knowledge graph and dig deep into the potential implicit knowledge in the graph. Specifically， the relationship path is first used， and the identity relationship is introduced to enhance the path obtained in the diagram to the same length. Then stack all paths， and use dynamic programming to model all the relationship subgraphs that share the same head entity at once. Finally， a query-dependent attention mechanism is used to select highly correlated edges. Experiments show that the MRR， Hits@1 and Hits@10 of the model are 76.2%， 68.53% and 92.04%， respectively， which can effectively infer new knowledge， realize the mining of lung cancer knowledge graph information， and achieve the effect of improving the knowledge graph of lung cancer diagnosis and treatment.关键词：knowledge graph;lung cancer;graph neural network;knowledge reasoning
- 24
- |
- 6
- |
- 0
发布时间：2023-04-18 - 摘要：Knowledge graph has received extensive attention in the field of recommendation， and it is often used as auxiliary information to be embedded in recommendation models to better alleviate the data sparsity and cold start problems of traditional recommendation algorithms. However， the input vector of some models is relatively sparse， and the feature interaction between users and items is not fully explored， which makes the representation between users and items less accurate and affects the performance of the model. Therefore， a knowledge graph recommendation model （BAKR） based on the fusion attention mechanism of FGCNN and MKR is proposed. First， FGCNN's Feature Generation module is used to extract feature vectors of users and items. Secondly， the knowledge graph is used to obtain the dependencies between entities， embed the implied auxiliary information into the model， and then redistribute the user's preference weight value through the attention mechanism to better assist the recommendation task and improve the recommendation performance. Finally， simulation experiments are carried out on the MovieLens-1M and Book-Crossing dataset， and the experimental results show that the accuracy of the model for the recommendation effect is significantly improved.关键词：recommendation system;knowledge graph;attention mechanism
- 20
- |
- 4
- |
- 0
发布时间：2023-04-18 - 摘要：At present， the image description model based on the coder-decoder framework does not consider the difference of the same word in different sentences at the encoding stage， and does not consider the hierarchical structure of the language sequence at the decoding stage. To solve this problem， the image description is studied based on deep learning， and the image description model based on Bert word vector and ordered memory network is designed by making full use of the visual features of the image and the text information of the reference sentences. The model adopts the coder-decoder framework. The encoder is responsible for obtaining the information of image and reference text， and the decoder is responsible for outputting the predicted text. The encoder uses the combination of Perception-v4 network and channel attention and spatial attention mechanism （CBAM） to obtain image features， and uses Bert model to vectorize the reference text to obtain the information of the reference text. Input the acquired visual features and text information into the decoder， use the strategy network and value network with good decision-making ability to provide guidance for the decoder， and generate the final image description sentence through the ordered memory network （ON-LSTM） combined with adaptive attention. This model is an effective image description model with an increase of 0.7%， 1.1%， 0.6% and 0.7% respectively on BLEU-1， BLEU-4， CIDEr and Meteor in the MS COCO Caption 2014 dataset compared with the basic model.关键词：BERT;ordered memory networks;image caption;deep learning
- 22
- |
- 3
- |
- 0
发布时间：2023-04-18 - 摘要：In order to solve the problem of slow convergence speed of arithmetic optimization algorithm and difficulty in finding local optimal value， an improved arithmetic optimization algorithm IAOA with updating strategy of optimization candidate solution is proposed. First， the update strategy of each dimension value of each candidate solution in the arithmetic optimization algorithm is changed to multiplication and division or addition and subtraction operations. Secondly， the position update strategy of the algorithm candidate solution is optimized to increase the global exploration capability and candidate solution diversity， so that the algorithm can find a better value. The simulation experiment of 21 benchmark functions shows that IAOA algorithm achieves the optimal value in 11 benchmark functions， and has faster convergence speed in most benchmark functions， compared with the Tartar sea squirt group algorithm， grey wolf optimization algorithm， Archimedes optimization algorithm and arithmetic optimization algorithm， in order to provide new ideas for solving arithmetic optimization algorithm.关键词：arithmetic optimization algorithm;swarm intelligent optimization algorithm;hyperparameter optimization;convergence speed;update policy
- 11
- |
- 5
- |
- 0
发布时间：2023-04-18 - 摘要：In order to meet the needs of multiple trajectory simulation of moving objects， a workspace solution and parameter optimization method for a four rope three degree of freedom planar parallel mechanism is proposed. Firstly， the static model is established by analyzing the structural characteristics of the mechanism， and the feasible workspace of the force screw under constant attitude is solved based on the orthogonal complement method. Then， the influence of structural parameters on the workspace size of the mechanism is studied， and the improved weight particle swarm optimization algorithm is used for optimization. Finally， compare and analyze the workspace size before and after optimization. The experimental results show that the number of workspace pose points of the optimized mechanism configuration is 27.5% higher than that before optimization. It is expected to provide a new solution for the optimization of mechanism workspace.关键词：cable-driven parallel mechanism;moving target simulation;workspace;orthogonal complement;PSO;parameter optimization;parallel robot
- 28
- |
- 6
- |
- 0
发布时间：2023-04-18 - 摘要：In order to explore the impact of connected vehicles on the overall road capacity， based on the characteristics of mixed traffic flow composed of connected automatic vehicles （CAV） and connected manual vehicles （CHV），analyze the eight types of vehicle driving and the spatial distribution proportion of the mixed traffic flow on the shared lane， and deduces the basic diagram model of the mixed traffic flow with different types of networked fleets on the shared lane. By changing the values of CAV permeability， right turn probability and the maximum size of vehicle queue in the formula， the traffic capacity of mixed traffic flow on the shared lane is comprehensively evaluated with average delay time and average fuel consumption. The simulation results show that under the environment of normal traffic flow density and high traffic flow density， with the increase of CAV penetration rate， the vehicle capacity and fuel consumption economy are increasing. When the probability of right turning vehicles is greater than 40%， it is necessary to consider adding special right turning lanes to improve the traffic flow performance. It is expected to provide reference for the study of road test of Internet of Vehicles.关键词：mixed traffic situation;vehicle platoon;shared lane;sensitivity analysis
- 12
- |
- 5
- |
- 0
发布时间：2023-04-18 - 摘要：In the process of SLIC clustering and segmentation， the redundant updating of pixel marks causes the algorithm to take a long time. In order to solve the above problems， a convex hull based SLIC super-pixel segmentation algorithm is proposed. Firstly， according to the gray characteristics of the image grid， the target region is filtered out by using the threshold mark. Secondly， the convex hull algorithm is used to construct the convex hull of the image and determine the initial clustering center based on its structural characteristics. Finally， based on the initial clustering center， SLIC algorithm is used to complete the image segmentation. In the MATLAB environment， the proposed algorithm and three related algorithms are tested. The experimental results show that the proposed algorithm can effectively improve the efficiency of the algorithm and has better segmentation quality.关键词：SLIC algorithm;superpixel;convex hull;clustering segmentation
- 9
- |
- 7
- |
- 0
发布时间：2023-04-18 - 摘要：The use of compressed sensing technology in earthquake monitoring and early warning system can greatly reduce the resource occupation. The measurement matrix is a key step in compressed sensing technology， and affects the hardware implementation mode and signal reconstruction effect. Therefore， it is necessary to construct a measurement matrix which is easy to be realized by hardware and has good reconstruction effect. Aiming at the reconstruction uncertainty of randomness measurement matrix and the complexity of hardware implementation， a compressed sensing measurement matrix applied to strong earthquake signals is proposed. The measurement matrix is constructed by the Logistic chaotic pseudo-random sequence mapped by the improved Logistic chaotic system. The proposed Logistic chaotic sequence measurement matrix meets the signal reconstruction conditions， and compared with the random measurement matrix， the simulation results show that the reconstruction error value of the measurement matrix is reduced， the time consumption is reduced， the reconstruction uncertainty is eliminated， and it is easy to be realized by hardware physics， so it has strong practical value.关键词：earthquake monitoring and early warning;compressed sensing;measurement matrix;chaotic system;earthquake signal
- 9
- |
- 2
- |
- 0
发布时间：2023-04-18 - 摘要：Due to the limited number of antennas in the radio interference array， the complete sky image information can not be obtained in the Fourier space， which makes the reconstructed image obtained by inverse Fourier transform of the measured data appear artifacts， resulting in the ambiguity of the details of the radio source.In order to further improve the quality of image reconstruction， adopt a radio image restoration scheme based on residual condition generation adversarial network and applies it to SKA simulation data. The model can recover the original image from the degraded image under undersampling after optimization training. Experimental results show that compared with the commonly used radio image clean algorithm， the proposed deep learning algorithm has a significant improvement in the image index PSNR and SSIM， and can better restore the brightness distribution map of celestial bodies.关键词：interference imaging;image reconstruction;SKA image restoration;radio observation;conditional generative adversarial nets;residual network
- 12
- |
- 4
- |
- 0
发布时间：2023-04-18 - 摘要：3D laser scanners on different platforms are suitable for different projects and have different characteristics. To study the differences between hand-held 3D laser scanners （HLS） and terrestrial 3D laser scanners （TLS） in building data acquisition applications， take a building about 8 meters high as the object，and uses the two instruments to acquire data respectively. Then the study completed point cloud data preprocessing and surface model reconstruction.Based on the processed 3D model of the building，the study extracts the side length of the building and slices point cloud.The study compares and analyzes the performance of the two scanners for obtaining point clouds and modeling results from three aspects：data detail expression quality，side length measurement accuracy and time efficiency of data acquisition，and quantitatively evaluates the two devices in these three aspects.The empirical study shows that hand-held 3D laser scanner has a significant time efficiency advantage in data acquisition， which can save 83% time compared with terrestrial 3D laser scanner and obtain more point cloud data；The quality of data collected by terrestrial 3D laser scanner is relatively high， the average roughness is lower， and the detail expression is better.Compared with the measured side length of buildings， the measured side length of 3D models constructed by the two kinds of scanners has high precision correlation consistency. The results of this research can be used for the collection and modeling of building point cloud data under different project requirements， as well as the application of the 3D laser scanner on which platform to choose， which can be used as a reference for related technical performance.关键词：terrestrial 3D laser scanner;hand-held 3D laser scanner;3D modeling;SLAM
- 17
- |
- 3
- |
- 0
发布时间：2023-04-18 - 摘要：For robust principal component analysis （RPCA） algorithm based on traditional rank function and sparsity function approximation is not high enough， present a new moving object detection based on the fractional norm and minimax concave penalty （MCP）. Firstly， the rank and sparsity functions are approximated by the fractional norm and the minimax concave penalty function to achieve a better approximation of the low-rank and the sparse parts respectively， so as to extract a better moving object. Then， the alternating direction method of multipliers （ADMM） is used to solve the proposed model. Finally to reflect the advantages of the proposed method， through simulation experiment， the average F-measure value of the proposed method is 0.695 76. Compared with other RPCA-based method ， the effect is better， and compared with the best performance comparison algorithm， F-measure value increases approximately 17%.关键词：nonconvex robust principal component analysis;minimax concave penalty;fractional norm;alternating direction method of multipliers;moving object detection
- 12
- |
- 2
- |
- 0
发布时间：2023-04-18 - 摘要：The cigarette case defect detection algorithm on the quality inspection machine is improved， and the objection detection algorithm is integrated according to the inherent characteristics of the cigarette case defect detection to improve the accuracy of the smoke case defect detection.The improved Faster-RCNN is used instead of the traditional template matching， and the 4K cigarette case image collected by the CCD line scan camera is used for cigarette case defect detection. By adding the FPN layer， the accuracy of cigarette case defect detection increased by 1.4%. By using involution in the FPN， by comparing the detection accuracy of cigarette case defects of different sizes of involution kernels， it is found that the detection accuracy of cigarette case defects using 5×5 involutional kernels is greatly improved， increasing by 1%， and the detection time of a single cigarette case using the model of internal convolution did not increase. Finally， the accuracy of cigarette case defect detection is further improved through transfer learning， reaching 97.9%.The results show that the proposed algorithm can realize the positioning and classification of cigarette case defect detection， and has a high accuracy rate.关键词：image processing;CCD line scan camera;cigarette case defect detection;inner volume convolution
- 17
- |
- 4
- |
- 0
发布时间：2023-04-18 - 摘要：Aiming at the problems of SURF algorithm in image matching， such as long operation time and low matching accuracy， DAISY descriptor and RANSAC algorithm are used to optimize accelerated robust feature （SURF）， and an image matching method based on improved SURF feature points is proposed. Firstly， taking advantage of DAISY descriptor's simple structure， low complexity and high matching accuracy， DAISY descriptor is used to replace SURF feature description operator based on SURF algorithm feature point detection， and RANSAC algorithm is used to delete mismatched points. The experiment shows that the algorithm is more robust than SURF algorithm in many complex situations， such as image blur， angle rotation， illumination change， JPEG compression ratio change， etc. It can improve the matching logarithm， eliminate the wrong matching points， reduce the running time of the algorithm， and the matching accuracy is more than 95%.关键词：feature matching;speeded-up robust features;DAISY descriptor;RANSAC
- 17
- |
- 9
- |
- 0
发布时间：2023-04-18 - 摘要：The shape of leaves usually has small inter-class differences and large intra-class changes in plant leaf recognition， which pose a great challenge to accurate identification，therefore an accurate leaf recognition method based on deep squeeze-and-excitation residual network is proposed. On the basis of the deep residual network， we firstly learn the importance of each convolution channel in the classification task by means of squeeze and excitation operations. Then， we rescale the weight of each channel according to its importance to promote or restrain the corresponding channel. The proposed method is evaluated on the classic leaf recognition dataset （Flavia）. The results show that the proposed method is superior to other classic methods in terms of recognition accuracy and robustness to noise. In addition， the recognition accuracy under complex conditions can reach up to 97.75%.关键词：leaf recognition;convolutional neural networks （CNNs）;residual;squeeze-and-excitation;rescale weight
- 12
- |
- 5
- |
- 0
发布时间：2023-04-18 - 摘要：To improve the efficiency of device-side face detection， a device-side face detection method based on YOLO pruning is proposed. Firstly， improve the YOLO network to realize face detection and face key point regression. Then， the model pruning strategy is used to prune the network model to reduce the model complexity. Finally， the robustness of the model is enhanced by integrating the multi-task loss function and training strategy. The experimental results show that， compared with the existing lightweight face detection algorithms on the face detection data set Widerface， the proposed method can improve the detection accuracy and reduce the model complexity. Meanwhile， it can also meet the face detection of the equipment demand.关键词：face detection;key point regression;model pruning strategy;comprehensive loss function
- 47
- |
- 7
- |
- 0
发布时间：2023-04-18 -
Structured Interactive Assignment Design from the Perspective of Knowledge Internalization Levels AI导读
摘要：Assignment design and practice are hot issues in the field of teaching research， and new connotations and standards have been derived from the environment of information teaching. In view of the lack of hierarchy and system in assignment design， this research defines the design concept of multi-level knowledge internalization to guide learners， summarizes the design of structured assignment， and elaborated on the characteristic elements of the assignment. On this basis， the article constructs a four-layer model of structured interactive assignment： basic implementation， accumulation and precipitation， consolidation and reflection， and expansion and exploration， then discusses the form of knowledge representation that progresses with the internalization level. Finally， the article selects the "sorting algorithm" in programming teaching to carry out unit assignment examples. The results show that structured interactive assignment can effectively enhance the internalization of learners' knowledge at different levels， and can provide reference for the overall coherent online assignment design.关键词：online assignment;design architecture;knowledge internalization;assignment characteristics;teaching interaction;knowledge representation;teaching practice- 10
- |
- 8
- |
- 0
发布时间：2023-04-18 - 摘要：To promote the reform of curriculum teaching and deepen the ideological and political course construction in computer specialized curriculum structure， achieve the education function of each specialized course and realize the fundamental goal of morality education， according to the “computer network” course teaching and ideological and political education in the perspective of multiple collaboration， from the knowledge， emotion， life and other aspects of college students， this paper deeply explores the combination of ideological and political elements and course teaching， and designs the implementation strategy of in-class and off-class multiple collaborative education by integrating flipped classroom and rain classroom modes， so as to provide ideas for promoting the ideological and political teaching reform of computer courses and improving the quality of computer professional talent cultivation.关键词：multiple collaborative;ideological and political teaching in course;flipped classroom;teaching reform
- 23
- |
- 12
- |
- 0
发布时间：2023-04-18 - 摘要：To improve the program design capability of professional talents majored in computer science， established the OBE based online programming platform， including code management， open-source coding， continuous integration， etc.， by using the open-source architecture and relevant techniques of GitLab. The platform was applied in various levels of programming practice courses of Yunnan University in recent years. Thus， the capability and interest of programming based solving of practical problems could be cultivated simultaneously. By this platform， the practice course schemes were improved， the course contents were enriched， and the construction of textbook and casebase were strengthened， to a great extent. This platform has promoted the course system reform of computer science major， as well as the improvement of learning scheme， contents and results. Some experience of this platform’s construction and practice could provide reference for relevant universities and majors.关键词：computer science major;program design capability;GitLab open-source architecture;online programming platform;practice education reform
- 8
- |
- 4
- |
- 0
发布时间：2023-04-18 - 摘要：The traditional course teaching pays attention to in-class， but ignores out-class， neglecting the cultivation of students’ active learning and positive thinking ability， which is still far from the new requirements of teaching quality of computer courses in the new era， and does not conform to the “student-centered” concept emphasized by the first-class course construction and engineering education certification. Inspired by the record of undergraduate course problem inquiry activities， a teaching mode of computer specialty courses based on problem circulation sheet with smaller time granularity is proposed. Taking software engineering course as an example， a complete teaching process of problem circulation sheet is demonstrated. The teaching mode uses the problem circulation sheet to link before-class， in-class and after-class， as well as connect the adjacent chapters， which can greatly promote students’ learning enthusiasm and improve teaching quality on the whole.关键词：problem circulation sheet;computer specialty course;teaching mode;first-class course construction;student-centered
- 10
- |
- 7
- |
- 0
发布时间：2023-04-18 - 摘要：Aiming at the problems of unsuitable traditional teaching evaluation and too complicated and difficult to quantify evaluation indicators in hybrid teaching， this paper proposes an evaluation model that integrates student self-evaluation， teacher evaluation and intelligent evaluation. First of all， the keyword clustering method is used to study the related literature of mixed teaching evaluation， and the research status is analyzed in depth； Secondly， a complete mixed teaching evaluation model is constructed， and the design of the model is described in detail. In the intelligent evaluation part， knowledge tracking is used to evaluate students' knowledge mastery； Finally， the application and prospect of the model are discussed. The practice shows that the implementation process of the model is relatively simple and flexible， which can effectively improve the teaching efficiency， and has good application effect. It can provide reference and reference for universities to carry out mixed teaching evaluation in various forms.关键词：mixed teaching;knowledge tracking;teaching evaluation;student self-assessment
- 12
- |
- 4
- |
- 0
发布时间：2023-04-18 - 摘要：In recent years， Web application systems are widely used， and the demand for quality assurance methods and tools for Web applications is increasing. Software testing is an important part of the software development process. It can ensure the quality and reliability of the system by finding the faults in the system. However， in the agile development process， traditional manual testing has been inadequate. In order to meet the version requirements of the rapid iteration of Web applications and reduce the testing costs， Web function automated testing came into being. This method can not only shorten the test cycle， but also improve the test efficiency， providing a guarantee for the quality of the Web system. First of all， this paper introduces the general process of Web function automation testing， and analyzes the challenges it faces in terms of test sequence generation， test data generation， execution time and test result analysis. Then， from the aspects of automatic generation of test cases， automatic execution of test cases and automatic analysis of test results， the classification research is carried out， and the research results， existing tools and main problems in various fields are summarized and analyzed. Finally， on the basis of recent research and analysis， discuss the development trend of Web automated testing frameworks and tools from the aspects of code-free and intelligent， in order to provide reference for the development of Web functional automated testing.关键词：Web application;functional testing;automated testing;test case generation;test case execution;analysis of test results
- 14
- |
- 3
- |
- 0
发布时间：2023-04-18 -
Analysis of the Current Situation and Trends of Domestic Serious Games Research Based on CiteSpace AI导读
摘要：Domestic serious game research has lagged behind the development of the industry. In order to visually analyze and display the overall research status and development trend in this field， the relevant literature in the field of serious game research in CNKI is visually analyzed and summarized with the help of the literature metrological analysis software CiteSpace. By analyzing the number of posts， researchers and institutions， research fields and hot spots， sort out the research status of serious games in China， and discuss the problems and hot trends faced at this stage. The practice shows that the research on serious games in China started relatively late， focusing on the basic concept and development process of serious games， design and dissemination of serious games， etc. The cooperation between researchers is not strong， and the basic theoretical research on user cognition， learning model， interaction mode， etc. is weak， lacking successful market strategies， and lag behind the development needs of the industry. And seek the breakthrough point of theoretical research and design practice from the aspects of integrating traditional culture， introducing new technology， balancing entertainment and learning， and paying attention to gender issues， in order to provide reference for the development of domestic serious games.关键词：knowledge graph;serious game;visual analysis;CiteSpace;smart education- 9
- |
- 2
- |
- 0
发布时间：2023-04-18 - 摘要：The correlation filtering algorithm has high efficiency in the process of target tracking. Inspired by the sub-block tracking algorithm， the idea of segmentation is introduced into the correlation filtering tracking framework. The correlation filtering improves the segmentation algorithm. Firstly， according to the correlation filtering tracking framework， combined with the correlation filtering theory， the development of correlation filtering is described. Then， it introduces the method of selecting the target block and how to determine the target location， explains the steps of the target block mode， and analyzes the three types of target block. Secondly， two types of blocking ideas are summarized， and the improved algorithm for kernel correlation filtering （KCF） target blocking is described in detail. In the overall performance evaluation， the success rate comparison of interference attributes of 6 blocking algorithms and the accuracy and success rate comparison between 3 blocking algorithms and 11 Correlation Filter Trackers （CFTs） algorithms are listed. It fully reflects the advantages of block method and large development space. Finally， in view of the existing problems and difficulties， the future development trend of the segmentation algorithm of relevant filters is proposed.关键词：correlation filtering;target tracking;kernel correlation filtering;block model;template update
- 15
- |
- 4
- |
- 0
发布时间：2023-04-18
- 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