最新刊期
卷 21 ， 期 12 ， 2022
- 摘要：Using EEG signal can objectively and continuously evaluate the mental workload level of human under language comprehension task and improve the overall performance of human-machine system. However， in practice， some EEG channels contribute little to mental workload recognition and only increase the computational complexity of machine learning models. Therefore， it is necessary to select channels for Multi-lead EEG. In this study， based on the wrapping channel selection algorithm， SMOTETomek strategy is used to oversample the training data of each channel EEG data， and then random forest training is used to obtain three predictive performance indicators of the classifier： the Macro F1 score， G-mean and AUC value. Finally， the contribution degree of channel weight is obtained by fusing and ordering the three indexes， and the redundant channels are deleted by step elimination method. Compared with the traditional filtering channel selection algorithms， this method achieves better classification effect by using least 9 channels. In addition， the location of the selected channels is consistent with the existing brain regions of language comprehension and mental workload， which verify the repeatability of the channel selection method.关键词：mental workload;channel selection;unbalance data;random forest;SMOTETomek;EEG
- 64
- |
- 18
- |
- 0
发布时间：2023-01-19 - 摘要：Aiming at the key technical problems and challenges existing in traditional facial expression recognition research， such as model training， application implementation and small sample learning difficulties， propose a lightweight facial emotion recognition method based on attention mechanism. In this method， the Xception network is improved by adding an attention module as an auxiliary block， and the model's attention is focused on the meaningful image input and the region with greater expression intensity through the channel attention module and the spatial attention module respectively. The Switchable Normalization （SN） is used to select different operations for normalization layers at different locations and depths in the neural network， giving greater weight to normalization approaches that benefit the network layer. Compared with other classical models on FER2013 dataset， the accuracy of this method can reach 86%， and the recognition rate of angry， happy， surprised and natural is higher， which can reach more than 90%. In addition， the model improves the traditional convolutional neural network for unconstrained facial expression scenes， and makes up for the defects of traditional convolutional neural network such as large number of parameters， difficulty in training， difficulty in landing， gradient disappearance and gradient explosion.关键词：facial emotion recognition;deep learning;attention model;lightweight network;auxiliary block
- 55
- |
- 18
- |
- 0
发布时间：2023-01-19 - 摘要：Deep learning technology of online attack and defense experimental courses has developed rapidly， and has been widely used in the task of recommending traditional Chinese medicine. Aiming at the problems of low recommendation accuracy and large model parameters of traditional neural network model in the application of traditional Chinese medicine recommendation， a lightweight drug recommendation method based on multi-level feature fusion is proposed. On the basis of the features of TextCNN model， such as less parameters and comprehensive feature extraction， the symptom semantic features and sequence features are further integrated， so as to obtain more comprehensive characteristics of symptomatic medicinal materials to complete the task of recommending traditional Chinese medicine， and it is verified on the public data set of traditional Chinese medicine. The experiment shows that the F5 score recommended by this method for medicinal materials reaches 0.2419， which is significantly improved compared with the baseline model， and the model size is only 4.26M. In addition， the ablation experiment was conducted to analyze the impact of different model components on the recommendation task， which verified the effectiveness of the proposed method， with a view to providing a new method for the recommendation of traditional Chinese medicine.关键词：herbs recommendation;machine learning;Glove;TextCNN;feature fusion
- 38
- |
- 12
- |
- 0
发布时间：2023-01-19 - 摘要：To improve the problem that the existing human pose estimation network is not effective in locating the human keypoints with diverse scales， a multi-scale bottleneck layer for feature fusion is proposed， and human pose estimation is performed based on this bottleneck layer. The multi-scale bottleneck layer discards the single large convolution often used in common feature fusion modules and uses multiple small convolutions connected in cascading hierarchical manner， which allows it to obtain multi-scale information at a finer grain level and iteratively perform feature fusion， thus giving the network a stronger multi-scale feature extraction capability to adapt to dynamic changes in the size of human keypoints. The experimental results show that the human pose estimation network based on multi-scale bottleneck layer achieves 76.2%， 73.0% and 81.2% of mAP， APM and APL， respectively， on the COCO dataset， which proves that the multi-scale bottleneck layer can improve the network scale robustness and thus effectively locate the human key points with diverse scales.关键词：deep learning;convolutional neural network;human pose estimation;multi-scale bottleneck layer;cascading hierarchical structure
- 33
- |
- 9
- |
- 0
发布时间：2023-01-19 - 摘要：In order to solve the problems of deep learning algorithm in traffic sign detection and recognition， such as large model parameters， weak real-time performance and low recognition accuracy， a traffic sign detection algorithm based on DPPC-YOLO is proposed. On the basis of YOLOv3 algorithm， the algorithm uses depth separable convolution（DSC） to reduce features to extract network parameters， introduces pyramid pooling module（PPM） and path aggregation network（PaNet） to optimize small target prediction results， and uses complete intersection over union （CIoU） ratio loss function to improve the positioning accuracy of prediction box， effectively improving the detection accuracy and real-time performance of the algorithm. The experimental results show that the average recognition accuracy of DPPC-YOLO algorithm is 89.73%， the detection speed is 11.8 FPS， and the model capacity is 49.5 MB， which is more accurate and fast than the traditional algorithm， so as to provide reference for intelligent identification of traffic signs.关键词：traffic sign recognition;YOLOv3;DSC;PPM;PaNet;CIoU
- 27
- |
- 8
- |
- 0
发布时间：2023-01-19 - 摘要：To further optimize the user experience of intelligent voice products， based on literature research and with the help of Donal. Norman ' s user experience theory， analyze the problems of user experience in the process of using intelligent voice products. Combined with the comprehensive analysis of the user experience principle and the development situation of intelligent voice products， the relevant experience improvement measures are proposed to promote the upgrading of intelligent voice products and improve the user experience. Through the study of user perception， service scene and service ecology in the process of using intelligent voice products， this paper puts forward some countermeasures to improve the user experience of intelligent voice products， such as multi-mode intelligence system， multidimensional collaborative experience and digital service ecology， which can effectively improve the user experience of intelligent voice products， so as to enhance the market competitiveness， user satisfaction and loyalty of intelligent voice products.关键词：voice interaction;user experience;intelligent product
- 20
- |
- 5
- |
- 0
发布时间：2023-01-19 - 摘要：Research on patent drawings in the mechanical field based on deep learning technology， fully explore and use patent labels information， seek supplementary methods for patent search， and propose a patent label recognition method CRNN_Eca based on improved CRNN algorithm. Change the feature extraction backbone network to ResNet34， and integrate the ECA module in ECA-Net to form the Eca-ResNet feature extraction network. The ECA module is an extremely lightweight and efficient channel attention mechanism. The original image is extracted by the Eca-ResNet network and then undergoes sequence conversion to generate the corresponding feature sequence. The result of the reference sign recognition is predicted and output through the deep two-way GRU network and CTC. The accuracy of the algorithm on the validation set and test set with reference signs reached 90.15% and 88.27%， which were 4.09% and 4.17% higher than the original CRNN algorithm， and the detection rate was greatly improved. Experimental results show that the CRNN_Eca algorithm for patent reference signs can achieve higher recognition accuracy and faster recognition speed， which is an effective patent reference sign recognition algorithm.关键词：patent part label;text recognition;attention mechanism;natural language process;deep learning
- 13
- |
- 4
- |
- 0
发布时间：2023-01-19 - 摘要：In order to improve the accuracy of feature screening and model prediction of anti-breast cancer drug candidates， propose a new method for screening multiple combinations of features for anti-breast cancer drug candidates——estrogen receptors alpha （ERα） molecular descriptors to perform feature screening， and to construct compounds based on the molecular descriptors screened Activity prediction （QSAR） model. In this study， the RMSE， MAE， and MAPE of the phase model of the 1DCNN algorithm are 0.40， 0.41 and 0.08， respectively. Compared with the traditional random forest and support vector machine algorithm， the prediction effect is improved by 10%. Feature screening methods based on multiple combinations and 1DCNN model prediction provide new ideas for feature screening and activity prediction of pharmaceutical compounds， which can be subsequently used for other pharmaceutical compound predictions.关键词：machine learning;breast cancer;hormone receptor alpha subtype;drug prediction;combinatorial feature screening;1DCNN
- 35
- |
- 6
- |
- 0
发布时间：2023-01-19 - 摘要：Bridges tend to deteriorate over time. In order to avoid life and property safety problems， identifying， predicting and analyzing the data monitored on the bridge structure is helpful to find the damage location and judge its damage type as soon as possible. Therefore， BP neural network is constructed to extract the characteristics of bridge monitoring data and identify and classify multi class data. Bayesian algorithm is used for hyperparametric optimization， and four machine learning algorithms after hyperparametric optimization， such as decision tree， neighbor algorithm， random forest and support vector machine， are used as control experiments. The experimental results show that the BP neural network has better performance， the prediction accuracy reaches 0.930 2， and its accuracy， recall and F1 index are also better.关键词：BP neural network;machine learning;damage identification;Bayesian optimization;hyperparametric optimization
- 25
- |
- 6
- |
- 0
发布时间：2023-01-19 - 摘要：In order to solve the problems of semantic missing and named entity nesting in Chinese medical named entity recognition task and improve the entity recognition effect in the clinical medical records of famous Chinese medicine， a named entity recognition model based on adaptive word embedding RoBERTA-wwm is proposed for the clinical medical records of famous Chinese medicine. The initial vector of the original text in the medical record obtained by the RoBERTa-wwm pre-training model adopts the Soft-lexicon method to dynamically fuse the dictionary information， perform vocabulary enhancement， and generate the text semantic vector. Random field （CRF） decoding extracts entities. The model proposed in this paper achieved an F1 value of 86.88% on the clinical medical record data set of the famous Chinese medicine Li Tiejun in the treatment of cardiovascular disease. Compared with the RoBERTa-wwm-CRF and Bert-CRF models， the F1 value was increased by 5.93% and 5.87%， respectively. There has also been an increase in speed. The adaptive word embedding is introduced into the conventional RoBERTA-wwm model for vocabulary enhancement， so that the model can better learn the textual semantic information. Compared with other baseline models， the named entity recognition task in the clinical medical records of famous Chinese medicine has significant advantages.关键词：information extraction;named entity recognition;famous traditional Chinese medicine electronic medical record;RoBERTa-wwm;vocabulary enhancement
- 8
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：For the noise reduction of underwater injury-containing noise-containing signals， a new method of taking wavelet packet analysis based on CEEMDAN decomposition for comprehensive noise reduction is proposed. Firstly， wavelet packet denoising is used to pre-process the initial noisy information， and multiple characteristic modal components are obtained by means of CEEMDAN decomposition， and the modal components containing dominant noise are selected by autocorrelation function； then the modal components of dominant noise are denoised by using progressive semi-soft threshold， and then the final pure signal is obtained by signal reconstruction. The feasibility of the noise reduction method is verified by denoising the classical test signal and the probing noise signal measured by the simulated underwater probing Mk4 small multi-echo measuring instrument. The method has the advantages of CEEMDAN decomposition and progressive semi-soft threshold synthesis， and the noise reduction effect is better compared with the general CEEMDAN and wavelet packet denoising methods， which can be applied to the noise reduction processing of similar noise-containing flaw signals.关键词：signal processing;underwater flaw detection;wavelet packet transform;CEEMDAN;noise reduction
- 24
- |
- 8
- |
- 0
发布时间：2023-01-19 - 摘要：In terms of the problem that the dominant manufacturer may ignore the benefit of other supply chain members when making decisions in traditional order allocation， it is proposed that decision-makers should make win-win decisions for supply chain members from the perspective of cooperation.Two important links in the supply chain are selected for joint optimization. Combined with the actual situation of enterprise， the mathematical model of order allocation associated with vehicle routing optimization is set up and is solved to use the double genetic algorithm with the goal of maximizing profits of all members including a single manufacturer and several distributors.Through case simulation， the validity of the model and algorithm is verified.The cooperation between supply chain members for the common goal will be conducive to play an important part in the fierce competition in the future.关键词：order allocation;vehicle path optimization;joint optimization;two-layer genetic algorithm
- 13
- |
- 6
- |
- 0
发布时间：2023-01-19 - 摘要：Aiming at the problem that whale optimization algorithm is easy to fall into local optimum and convergence speed is slow， a whale optimization algorithm based on extremum individual guidance is proposed. Firstly， the parameter C is dynamically adjusted in a nonlinear way to balance the global search capability and local search capability. Surrounded by then， in feeding stage， the optimal individual USES group and neighborhood search for the optimal individual leads whales， based on the current position， by the whales of the individual to the group of optimal whales swimming at the same time， in the neighborhood of the optimal individual form small range around together， enhance the global search ability of algorithm， to prevent the algorithm falls into local optimum. Thirdly， in the stage of searching prey， the optimal individual in the neighborhood is used to guide whales to search， which improves the convergence speed and optimization accuracy of the algorithm. Finally， 23 test functions including single peak， multi-peak and fixed dimension are selected to test the improved algorithm. The simulation results show that the improved algorithm has higher optimization accuracy， and the rank sum test results have significant differences， that is， the improved algorithm has better optimization performance.关键词：whale optimization algorithm;extreme individual guidance;dynamic adjustment
- 18
- |
- 4
- |
- 0
发布时间：2023-01-19 - 摘要：Measuring the importance of nodes in networks is of great help to improve the robustness and reliability， but some classical centrality indices inevitably have limitations and one-sidedness. Therefore，propose a method that integrates the global and local characteristics of the network， which is more exact than the node deletion method. It defines the local importance of nodes by quantifying the similarity of node neighbors’ topological. The global importance of nodes is decided by the network connectivity after removing nodes' incident links. To validate the effectiveness and applicability of the proposed algorithms， contrast experiments were carried out in different kinds of networks. The results show that the method can effectively identify the vital nodes in networks. Compared with the node deletion method， the fineness of this method is improved by 16% in the Jazz network. This method optimizes the node deletion method and overcomes the disadvantage of node deletion method in distinguishing the importance of different "bridge" nodes. At the same time， nodes with the same connectivity after removal can further subdivide the importance according to the topology similarity of their neighbors.关键词：complex network;spanning tree;node similarity;topological structure;vital node
- 9
- |
- 5
- |
- 0
发布时间：2023-01-19 - 摘要：Aiming at the characteristics of high-dimensionality and class imbalance of DNA microarray data，and in order to improve the classification effect， an algorithm based on the combination of Spectral Clustering and feature selection oversampling on Minkowski Distance （SCFSO-MD） is proposed. SCFSO-MD uses spectral clustering to cluster features， defines feature discrimination and feature independence respectively， and uses the product of them as the basis for representative feature selection construct feature subsets. Feature discrimination is measured by feature standard deviation， and feature independence is measured by Spearman coefficient. In SMOTE oversampling， the Minkowski distance metric method under different orders is used to calculate the K nearest neighbor samples. The experimental results show that the F-value of the SCFSO-MD-S algorithm on the SRBCT， Leukemia， Colon Tumor and Ovarian Cancer datasets can reach up to 0.99，0.95，0.90 and 0.97， respectively. When the number of optimal feature subsets is still large， the effect of the SMOTE algorithm based on the Manhattan distance metric is more robust than the classification effect obtained by using the Euclidean distance.关键词：DNA microarray data;high-dimensional data;imbalanced dataset classification;spectral clustering;correlation coefficient;Minkowski distance
- 12
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：In order to solve the problem that the effectiveness evaluation of the depth learning level is difficult to achieve the desired results， a framework for evaluating the effectiveness of students' depth learning level based on situational awareness is proposed. This framework introduces the situational factors in the students' environment， and further explores the relationship between students， their situations and the characteristics of effectiveness evaluation of students' in-depth learning. At the same time， the framework uses trigonometric functions to adjust the hybrid leapfrog algorithm， avoiding the problems of local optimization and slow convergence speed， thus improving the ability of dimension reduction and prediction generalization in feature selection. The experimental results show that the effectiveness evaluation framework of students' deep learning levels based on situational awareness can be effectively applied to the prediction of students' learning effectiveness tendencies， thus providing a strong early warning and timely intervention for college students' teaching management.关键词：situational perception;framework for evaluating the effectiveness of deep learning level;teaching management;alarm;intervention
- 10
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：When the relationship between user notification interaction behavior and motion state is poor， the user experience will be reduced， and the traditional face-to-face research method is difficult to meet today's needs. Therefore， with the help of context aware user data collection tool CAUX， the data is processed automatically and analyzed qualitatively and quantitatively. After 8 months， the notification task model is basically established， and the research method about the relationship between user notification interaction behavior and motion state is explored from each stage of the model， and the method is applied to practical cases to explore the relationship between the two. The experimental results show that this method has advantages in informing the number and importance of discovery of the relationship between interactive behavior and motion state， improving the analysis efficiency of researchers， and has unique practical value. This method provides a new idea for studying the relationship between user notification interaction behavior and motion state.关键词：notification interaction behavior;motion state;data collection tool;qualitative analysis;quantitative analysis
- 9
- |
- 2
- |
- 0
发布时间：2023-01-19 -
Design and Implementation of Blockchain Traceability System for Sweet Potato Quality and Safety AI导读
摘要：Sweet potato is rich in variety resources and has many production links. Once there is a problem in the production process of sweet potato， it will not be able to accurately track the source of the problem. The actual demands of sweet potato from production to sales was analyzed， and the sweet potato quality and safety traceability system based on the blockchain was built to solve the shortcomings of the existing traceability information technology. By scanning the two-dimensional code label generated by the system， the clients can trace the whole process of "seed preservation - seed virus elimination - seed planting - sweet potato sales" of sweet potato dynamically.Meanwhile， by using this system， we can standardize the standard of planting sweet potato by enterprises， improve the quality of sweet potato， strengthen the ability of food safety supervision and enhance the brand value， etc. ，thus provide a valuable approach for the traceability of sweet potato， which has a great significance for sweet potato industry.关键词：traceability system;blockchain;sweet potato quality;quality safety;two-dimensional code- 23
- |
- 6
- |
- 0
发布时间：2023-01-19 - 摘要：On the future battlefield， unmanned ground vehicles can play an important role， including reconnaissance， strike， transportation， etc. In the face of the increasingly complex international situation and the uninterrupted military confrontation in local areas， the real battlefield environment will bring huge threat to the safety of soldiers' lives. At present， many countries are actively developing the unmanned ground vehicle command and control system in the battlefield environment. The medium-sized unmanned ground vehicle is suitable for urban or field assault reconnaissance， cover， transportation， strike and other ground scenarios. The design of the command and control system for unmanned vehicles will be based on QT and Bigmap map API， and the modules of the command and command system will be controlled based on the cooperative control theory. The command system and the unmanned vehicle will communicate with each other through UDP communication protocol. Interaction to achieve command control of unmanned ground vehicles.关键词：unmanned ground vehicle;command and control system;letter of agreement;collaborative control;information exchange
- 24
- |
- 8
- |
- 0
发布时间：2023-01-19 - 摘要：In order to realize the application of vortex beams in miniaturized devices， a cylindrical hole grating etched silicon microring resonator structure with the device radius of only 4.1μm is proposed. The expression of the output optical field is obtained by establishing a theoretical model with dipole points to represent the radiation scattered by grating elements along the inner wall of the microring resonator. The modeling results verify that the designed structure can selectively generate vortex beams of various topological numbers at different wavelengths. In addition， through the performance analysis and parameter optimization of the device. The research results show that the radiation loss of the microring will be greatly increased if the width of the microring waveguide is too small， when the width of the microring waveguide is 450nm and the coupling spacing is 100nm， the output power around the wavelength of 1 579μm is 16.3 dB， and the quality factor Q can reach as high as 15 800， which provides some valuable references for subsequent experimental tests and practical applications of vortex beams in free-space optical communication， optical imaging， sensing， and atomic magnetometer light source systems.关键词：FDTD;vortex beam;emitter;microring resonator;gratings;integrated optics devices
- 6
- |
- 5
- |
- 0
发布时间：2023-01-19 - 摘要：In order to improve the intelligence of the restaurant delivery system， relying on the ROS mobile robot platform of NVIDIA Jetson Nano b01， the AI delivery software system for indoor restaurants is designed. Based on the lightweight deep learning framework Darknet， the system is programmed with C + + and Python. It includes four modules： controller， voice， motion control and AI visual processing. Start the food delivery robot by voice， and after recognizing the QR code of customer ordering information， navigate to the target table to deliver food. At the same time， people along the way of food delivery are visually recognized based on YOLOv3 tiny deep learning， and welcome words are broadcast by voice when customers are recognized. The experimental results show that the system can effectively complete the meal delivery task， the two-dimensional code recognition time is about 0.003s， and the character recognition accuracy is up to 99%. The system can automatically start and broadcast voice， and the design scheme is reasonable and effective， which can provide reference for the AI food delivery field.关键词：artificial intelligence;food delivery system;Jetson Nano platform;robot operating system
- 14
- |
- 4
- |
- 0
发布时间：2023-01-19 - 摘要：Currently， most city buses in China implement the “one-card system” mode， which makes the passenger's alighting data unable to be recorded in the data field. Therefore， the prediction of the passenger's alighting point has become a current research hotspot. However， in the actual operation of buses， it is not uncommon for the replacement of hardware equipment， real-time route changes， and the situation where the driver randomly changes the driving route. As a result， the subsequent prediction errors for passengers getting off the bus are relatively large， which affects the overall planning of urban public transportation by the management department. How to verify the accuracy of the previous data sources is an urgent problem to be solved to promote the application of bus data informatization. Taking Changsha City as an example， This article relies on the Changsha City Transportation Comprehensive Operation Coordination and Emergency Command Center System to solve the problem of field mismatch in the system. In this regard， this paper proposes a heterogeneous data matching method for urban public transportation from a technical point of view. The method first finds out the corresponding relationship of different fields at the time point， and obtains the initial matching rate by using threshold analysis to integrate the bus IC card swiping data and the bus GPS arrival and departure data； then， the unmatched vehicle information is reported as the experimental result Relevant departments conduct data review and order them to feed back the correspondence table of vehicle relations for the next month； finally， by summarizing the correspondence table fed back from the previous month with the experimental results of the current month， a complete vehicle relation correspondence table is finally obtained. Five months after applying the strategy of this article to the public transportation system， the matching rate of public transportation data in Changsha City increased from 61% to 86%.关键词：public transportation;data matching;heterogeneous data;IC card data;GPS data;TOCC
- 14
- |
- 4
- |
- 0
发布时间：2023-01-19 - 摘要：Guangdong tobacco commercial system has built marketing， monopoly， internal management， logistics， finance and other application systems， and has accumulated a large amount of business data. How to effectively use the data and fully release the value of the data， and then help enterprise business innovation， promote digital transformation， and achieve quality and efficiency of state-owned enterprises， is a major issue for the high-quality development of the industry. In this paper， we analyze the current digital dilemma faced by tobacco enterprises' customer service and find that two sets of computing platforms need to be established to handle real-time streaming and offline batch computing respectively when dealing with data governance tasks such as data specification and data reuse. To address the current problem， a stream-batch computing architecture that is compatible with both real-time streaming and offline computing is used to design a data governance solution suitable for Guangdong tobacco business system， which enabled the application of 360-degree customer panorama and violation prediction and successfully improved customer service.关键词：stream-batch integration;data management;data specification;reuse capability;customer service
- 8
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：Under the background of the overall national security ， cultivating college students' network security concept is one of the important tasks to maintain national network security. At present， the content of college students' cybersecurity popularization education is single， fragmented， and lacks structured knowledge. In order to solve the above problems， propose to build a relevant knowledge graph to integrate cybersecurity popularization knowledge to help students master the systematic cybersecurity knowledge system， and to build a question answering system based on the knowledge graph to assist students in learning related knowledge and solving related problems. First introduce the knowledge map and the overall idea of building a knowledge map of network security popularization， and then build the overall framework of the knowledge question answering system on this basis， so as to help improve the popularization of network security education and help college students establish a good awareness of network security.关键词：overall national security;college Students' network security science education;knowledge graph;question answering system
- 10
- |
- 6
- |
- 0
发布时间：2023-01-19 - 摘要：In process of gray image enhancement， details are always lost and unnecessary noise is easily generated，propose a Riemann-Liouville fractional differential image enhancement algorithm based on Laplacian pyramid. Firstly， the traditional Riemann-Liouville fractional differential operator is improved and its pixel value is strictly controlled between 0~255， which greatly reduces the distortion. Then， to further enhance the image enhancement， the Laplacian pyramid is applied to enhance the original image layer by layer. Given that its strong protection of details and obvious suppression of noise， the algorithm is also effective in medical ultrasound images. It can be observed from the experimental results that this method not only improves the contrast of the image and the edge information of the ultrasound image， but also highlights the target without producing obvious noise so that the image details to the greatest extent can be preserved. It has a great prospect in gray image especially in ultrasonic image field.关键词：Laplacian pyramid;fractional differential;gray image;ultrasound image;image enhancement;noise suppression
- 15
- |
- 4
- |
- 0
发布时间：2023-01-19 - 摘要：Traditional air pollution concentration visualization methods cannot display complete visual results， and easily lead to visual perception problems， such as occlusion or chaos or even artifacts. In view of the above problems， an efficient visualization method is proposed. Based on the programmability of graphics hardware， the regional characteristics of atmospheric wind field are statistically analyzed， seed points are pre sampled， field lines are quickly constructed through tetrahedral point positioning algorithm， and the diffusible movement of pollutants is analyzed by combining arrow tracking and iso-surface rendering. At the same time， field line extraction is realized by clustering algorithm， and lighting model is designed to reduce the mutual occlusion between streamline and iso-surface， which effectively solved the perception problems in the three-dimensional flow field， shows a clear flow field mode， air pollutant concentration distribution and regional changes， and greatly improves the readability of air pollutant concentration data.关键词：multivariate visualization;GPU;atmospheric pollution;streamline;iso-surface
- 12
- |
- 5
- |
- 0
发布时间：2023-01-19 - 摘要：Fast target detection and instance segmentation technology for complex traffic images is the key to achieve autonomous driving. The existing two-stage method performs the detection and segmentation task in multiple steps， which takes a lot of time and is difficult to optimize. The one-stage method has a lot of post-processing after target positioning， which is difficult to meet the real-time requirements. To solve the above problems， propose a fast instance segmentation model （FISAT） based on target detection and generative adversarial network. Firstly， the parallel branch generation mask is introduced into the target detection network to segment each traffic instance object. Secondly， ROI class loss is added to learn each class mask， and perceptual loss is used to save mask image information. Finally， spectral normalization is applied to solve the slow convergence problem in the training process of generative adversarial network. On the benchmark of MSCOCO， the frame per second （FPS） of FISAT can achieve 47.0， which is five times that of MNC and FCIS. In terms of segmentation optimization， the Darknet extractor FPS achieves 43， which is 8.0 higher than that of the Resnet extractor. The average accuracy （AP） of FISAT is 7% higher than that of the two-stage Mask-RCNN and 24% higher than that of the one-stage method.关键词：autonomous driving;instance segmentation;object detection;generative adversarial network
- 7
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：The traditional fusion detection methods can not solve the problems of false detection and missing detection caused by redundant information in bimodal fusion. In order to make more effective use of bimodal information， a bimodal feature fusion pedestrian detection network （IWFC Net） based on light perception and convolutional block attention module is proposed. Firstly， the light perception value is extracted from the visible light image and used as the fusion weight to guide the feature fusion of bimodal images； Secondly， the attention mechanism of convolution block is strengthened by strengthening the fusion features， and the feature representation is strengthened； Finally， the enhanced fusion features are sent to the YOLO target detection layer for multi-scale pedestrian prediction， and the multi-scale pedestrian detection results are obtained. The experimental results on KAIST and LLVIP multispectral datasets show that the proposed method can not only effectively fuse the complementary information of dual mode images to improve the target detection accuracy， but also significantly reduce the rate of missed detection and false detection caused by redundant information， and can also achieve the detection speed of about 35 FPS， in order to provide a new solution for pedestrian detection.关键词：visible and infrared images;dual-mode image fusion;pedestrian detection;illumination aware;convolutional block attention module
- 12
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：Aiming at the situation that Gaussian mixture model is sensitive to the initial value and easy to fall into the local optimal value， a Gaussian mixture model image clustering method based on the improved EM algorithm is proposed. The method first uses the particle swarm with the inertial weight to increase and then decrease to initialize the parameters of the Gaussian mixture model， then introduces the approximate skeleton theory to improve and optimize the EM algorithm， solve the parameters of the Gaussian mixture model， and finally perform simulation experiments in the image clustering application. SSIM values of gaussian mixture model image clustering method based on improved EM algorithm are 6.31%， 4.20% and 1.38% higher than those of standard EM， RSEM and PSOEM， respectively. The cosine distance is 4.12%， 2.69% and 0.94% higher than that of which. Analysis of experimental results shows that the method can effectively improve the clustering effect of local pixel regions and obtain a clearer output image with cluster boundaries.关键词：PSO;approximate backbone;initialization;EM algorithm;image clustering
- 9
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：In the past 20 years， open source technology has developed rapidly and has deeply affected everyone's daily life， such as Android mobile operating system， Tensorflow， Caffe artificial intelligence deep learning platform， Hadoop， OpenStack， Docker cloud computing big data technology. At present， Chinese enterprises have a high degree of acceptance of open source technology， and the use of open source technology has become the mainstream. Spontaneous open source projects cover the whole stack of technical fields， and gradually move to the international forefront. As a result， the society has a huge demand for open-source talents and a serious shortage of open-source talents. Talent cultivation cannot be separated from education. However， the ability of open-source talents is different from that of traditional education system. In order to better build the soil for cultivating open-source talents， through the analysis of the development of open-source education at home and abroad in recent 20 years， the existing problems of domestic open-source education are analyzed and the future open-source education is considered.关键词：open source;open-source education;open-source software;maker education;college education
- 13
- |
- 6
- |
- 0
发布时间：2023-01-19 - 摘要：The practical courses of computer majors have problems such as low difficulties， single coverage of knowledge points， and lack of inspiration. It is difficult for students to effectively improve their programming ability and problem-solving ability. In order to enhance students' interests in learning， NPC game are introduced， and a set of teaching programs in accordance with aptitude that integrates interest cultivation， knowledge expansion， sufficient work， and teamwork are proposed. Compared with traditional courses， the effect of programming practice in the last three years verifies the effectiveness of the teaching reform of computer professional practice courses.关键词：computer major;practical courses;computational complexity;NPC games;interest driven
- 12
- |
- 9
- |
- 0
发布时间：2023-01-19 - 摘要：Benefiting from the continuous innovation of the Internet and information and communication technology， “Internet+” has become the trend of developing frontier intersection disciplines under the background of new engineering. Considering that the curriculum of information engineering major is limited by the traditional teaching mode， and it is difficult to stimulate students' innovation ability and patriotism， taking the course "Introduction to Energy Internet" as an opportunity， this paper puts forward a "three types and one line" hybrid gold course reform plan integrating ideological and political elements： Based on the teaching syllabus， it designed a teaching reform framework with appropriate ideological and political elements as the thread throughout， and made full use of the complementary advantages of the three modes of network cloud， offline teaching and flipped classroom to branch and intuitively demonstrate the key and difficult knowledge of the course. At the same time， it added the mutual evaluation mechanism of group speech. Based on the teaching syllabus， it designed a teaching reform framework with appropriate ideological and political elements as the thread throughout， and made full use of the complementary advantages of the three modes of network cloud， offline teaching and flipped classroom to branch and intuitively demonstrate the key and difficult knowledge of the course. At the same time， it added the mutual evaluation mechanism of group speech. Through specific cases， it is verified that the proposed educational reform program has more effective teaching and assessment results than the existing programs.It also can be better cultivate innovative talents for the Internet and energy， two national pillar industries.关键词：information engineering major;energy Internet;ideological and political education;hybrid golden course;TCP/IP
- 113
- |
- 21
- |
- 0
发布时间：2023-01-19 - 摘要：In order to further cultivate students' patriotism and improve computer students' awareness of network security and responsibility， teachers should actively explore the ideological connotation of network attack and defense courses， organically integrate ideological and political education with network security knowledge points in the course of teaching， and guide students to establish a correct world outlook， outlook on life and values through ideological and political education， so as to improve students' professional quality. Taking the network attack and defense course as an example， this paper puts forward how to integrate ideological and political education methods in the abstract network security knowledge points， so that students can master the network security knowledge， establish the feelings of network security family and country， further effectively improve the teaching effect and teaching quality of teachers， which can provide reference and reference for other courses to integrate ideological and political elements.关键词：network attack and defense;network security;curriculum ideology and politics;course reform;experimental teaching research
- 18
- |
- 3
- |
- 0
发布时间：2023-01-19 - 摘要：Carrying out ideological and political education in curriculum is an important measure to implement the fundamental task of establishing morality and cultivating people， and to promote the comprehensive reform of "three qualities" education. study stress has brought the students with depressive anxiety. Therefore， the ideological and political curriculum should not only lead the value， but also play a role in inspiring the spirit. Taking the C++programming course as an example， this paper first proposes the overall ideological and political teaching design of the course， and expounds how to integrate ideological and political elements into the teaching objectives； Then， taking the development and characteristics of C++programming as the teaching unit， and taking the epidemic situation as the clue， we designed ideological and political cases， so as to deepen students' understanding of C++object-oriented thinking， and finally achieved the goal of educating people by combining knowledge teaching， value guidance and spiritual motivation关键词：C++ programming;curriculum ideology and politics;case design;post-COVID-19 era;object-oriented
- 13
- |
- 7
- |
- 0
发布时间：2023-01-19 - 摘要：In experiment teaching for network establishment and management， real experiment plays an unsubstitutable role， while virtual simulation experiment can effectively solve the realistic problem of "unable to offer" of comprehensive， design and innovative experiment. Therefore， only the virtual-real combination can create the best experimental teaching environment for students. Based on this， after years of practice and exploration， this paper puts forward the application strategy of virtual and real experiment in the teaching of network establishment and management， the construction method of virtual simulation experiment environment and resource development strategy， and the design and implementation methods of experimental teaching projects. Practice has proved that the experimental teaching of virtual-real combination can significantly improve the teaching quality and effect.关键词：network establishment and management;real experiment;virtual simulation experiment;virtual-real combination
- 10
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：Program design competition is an important means for colleges and universities to strengthen talent training，this paper puts forward an effective model to help local universities cultivate innovative talents through computer programming competitions. According to OBE， on the basis of analyzing the characteristics of important competitions， the level of students and conditions in local colleges， a programming competition mode suitable for local colleges and universities has been built， the teaching mode and competition team have been improved，and the improvement of the organizational structure and team building， a competition organization and training mechanism has been formed. The students have made breakthroughs in ACM-ICPC and other important competitions. The students' competition performance is outstanding among similar institutions. The competition also promoted the teaching reform of the program design curriculum group， steadily improved the teaching quality， and achieved remarkable talent training results. Practice has proved that this model is an effective means and an important way to cultivate innovative talents in local colleges and universities.关键词：computer program design competition;Cultivation of innovative talents;teaching reform
- 10
- |
- 3
- |
- 0
发布时间：2023-01-19 - 摘要：Facing the integration method of ideological and political education in digital communication courses for the construction of new engineering， put forward the concept of "four ideological and political integration" from the aspects of teaching content， experimental courses， teaching mode， and academic frontier introduction. Moreover， the method of "three ideological and political introduction" is adopted in the implementation process. The implementation plan of ideological and political education in the whole teaching process is introduced. The feedback of classroom effects and the results achieved by students show that after integrating ideological and political， students' patriotism and belief in innovation and power are constantly stimulated， Both their interest in learning and learning effect are significantly improved.关键词：digital communication;ideological and political education;FPGA;new engineering construction
- 12
- |
- 3
- |
- 0
发布时间：2023-01-19 - 摘要：In view of the strong theoretical and practical characteristics and rapid development of computer science majors， as well as the deficiency that the training mode of master's degree students is not closely integrated with the actual needs of enterprises， propose a "problem-oriented + three docking type of industry-university-research collaborative education" model. By building a practical platform with university-enterprise cooperation， facing the problems of enterprises， taking the project as the grasp， establishing the joint cultivation program of graduate students， and promoting the collaborative innovation education of industry-university-research through the quality supervision and guarantee system of multi-level evaluation， the cooperation with Guangzhou NC Company shows that the education effect is remarkable and has important practical significance for the cultivation of computer engineering applied graduate students.关键词：computer science graduate students;industry-academia-research collaboration;cultivation model;problem-oriented;dual mentorship
- 11
- |
- 5
- |
- 0
发布时间：2023-01-19 - 摘要：The OBE teaching mode has been valued by many universities， and has gradually developed into a mainstream education mode. Taking the course of microcomputer interface technology as an example， taking the application skills， project development ability and professional ethics as the goal of talents， innovating the teaching content， increasing the proportion of practice， and comprehensively docking with extracurricular competitions. Attach importance to the ideological and political courses， and improve the moral quality of talents. The curriculum reform has effectively enhanced the enthusiasm and initiative of students.关键词：OBE;microcomputer principle and interface;Emu8086;Proues
- 14
- |
- 5
- |
- 0
发布时间：2023-01-19 - 摘要：Monocular vision SLAM is widely used because of its small size， low power consumption and rich information acquisition. In order to analyze the advantages of monocular vision SLAM in depth. Firstly， the basic principle of SLAM based on monocular vision is briefly introduced， and the key technologies of SLAM are summarized from the aspects of feature point detection， camera pose estimation， key frame selection， map creation， map and camera pose optimization， and closed-loop detection. Then， based on the feature method， direct method and hybrid semi direct method， the design framework of the current mainstream monocular vision slam system is introduced in detail， and more than 20 popular system performance characteristics and applicable scenarios are analyzed. Finally， the application of deep learning in camera attitude estimation， map creation and closed-loop detection is introduced， and compared with traditional methods to discuss the development trend of monocular vision SLAM.关键词：computer vision;monocular vision;SLAM;synchronous location and map creation;visual odometer
- 24
- |
- 12
- |
- 0
发布时间：2023-01-19 - 摘要：The task of texts semantic matching aims to calculate the semantic similarity of the texts， and then to measure the matching degree of two different texts. As an important foundation in natural language processing， texts matching can be used for numerous NLP tasks， such as textual implication， information retrieval， dialogue systems， machine translation， etc. The application of deep learning enables the texts semantic matching model to fully mine the texts semantic information and achieve better results in matching tasks. As a result， a variety of semantic matching models are reviewed based on different semantic composition structures， and the application of pre-trained language models in semantic matching are also introduced. In view of the diversity of texts matching tasks and original texts， the development of long texts matching tasks and Chinese semantic matching are also introduced， and representative deep semantic matching models are summarized. Finally， combined with the downstream practical application， the future development direction of texts semantic matching is discussed.关键词：text matching;natural language processing;deep learning;text semantic information;pre-trained language model
- 14
- |
- 33
- |
- 0
发布时间：2023-01-19 - 摘要：With the development of machine learning and deep learning models， artificial intelligence is being applied to the medical field. In oncology， the application of artificial intelligence in diagnosis and evaluation of medical images and genome data omics analysis has been increasing. Artificial intelligence can mine the law of development of things in big data through machine learning， find medical information that can be used， and generate new data， which will help medical workers more objectively understand and master the law of development of various diseases. Endometrial cancer is one of the three major malignant tumors in the female reproductive system， and its incidence rate is rising continuously. There are more and more reports about the application of artificial intelligence in endometrial cancer. Reviewing and analyzing these studies， it is found that the research mainly focuses on imaging diagnosis， and has achieved initial results in predicting deep muscle layer invasion， lymphatic vessel space invasion， lymph node metastasis， etc.， but few research results can be applied in clinical practice， and further exploration is needed in the future.关键词：artificial intelligence;endometrial cancer;machine learning;deep learning;neural network
- 14
- |
- 4
- |
- 0
发布时间：2023-01-19 - 摘要：RDMA technology is a direct memory access technology， which can realize the direct transfer of data from one computer to another without the intervention of the operating system， thereby realizing high-throughput and low-latency network communication. In recent years， in order to give full play to the advantages of RDMA and improve scalability， RDMA is usually combined with various technologies and is widely used. The combination of RDMA technology and database technology can greatly improve the high availability performance of the database； new RDMA-based storage technology can improve network performance， Storage expansion capacity， improve ease of use. On the basis of introducing the principles of RDMA implementation， aiming at the latest research progress in the application of the current RDMA technology industry， analyzing the specific ideas and existing advantages of the research， and exploring the direction in which RDMA can be further expanded in the future.关键词：RDMA;distributed storage;database;novel storage
- 12
- |
- 8
- |
- 0
发布时间：2023-01-19 - 摘要：Compared with cardiac ultrasound， magnetic resonance imaging and other diagnostic methods， cardiac auscultation has the characterisitic of high speed and low cost. The diagnosis of valvular disease by heart sound needs large clinical experience， long training period of a physician. However， the results are subjective. With the development of computer technology and machine learning algorithm， the use of machine learning model to complete the diagnosis of valvular heart disease has attracted more and more attention. The main steps of machine learning classification of heart sound signal are data acquisition， signal preprocessing， feature extraction and model training. This paper will introduce the main heart sound database and the steps of heart sound signal classification. Then summarize the effects and advantages of classification algorithms. Finally， the future will be prospected.关键词：heart sound signal;machine learning;neural network;signal segmentation;feature extraction;classification algorithm
- 27
- |
- 2
- |
- 0
发布时间：2023-01-19 - 摘要：In order to reveal the theme structure and evolution trend of the educational application of natural language processing technology. Based on the paper data of BEA workshop held by NAACL， an international top conference in the computer field， from 2003 to 2021， the LDA topic model is used for semantic analysis， and five topics are automatically identified， including intelligent teaching system， reading support， grammar error detection， mother tongue recognition， and text automatic scoring. Based on the recognition results， the topic intensity， the topic novelty and the topic evolution process are analyzed， and the research hotspots of the educational application of natural language processing technology are deeply explored. It is found that enriching application scenarios to promote the development of natural language processing technology， data-driven model and algorithm research， deep learning model and text representation methods are the key contents of researchers， with a view to providing reference for industry personnel to grasp the latest research trends.关键词：natural language processing;educational applications;LDA topic model;research hotspots
- 7
- |
- 3
- |
- 0
发布时间：2023-01-19
- 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