最新刊期
卷 23 ， 期 1 ， 2024
- 摘要：Regarding the research on electric excitation synchronous motor （EESM） in the field of vehicle motor control， a finite element model of EESM was designed using Maxwell， and a motor control system model was built based on Simulink. Then， based on the motor characteristics and electromagnetic torque analysis of the salient pole EESM， the maximum torque to current ratio （MTPA） vector control method was adopted in the control system to conduct joint simulation verification of the EESM control system in Simulink Maxwell. At the same time， the current loop of the control system applies linear active disturbance rejection control （LADRC）， and a model auxiliary extended state observer （MA-LESO） is designed in LADRC based on known motor parameters to observe and eliminate system disturbance variables. Simulation experiments show that the motor using MTPA vector control reduces speed fluctuations and enhances load capacity when loaded； The use of MA-LESO's LADRC controller in the current loop can reduce the fluctuation of motor speed in steady state， enhance anti-interference performance under load， and ensure relatively stable motor operation； The waveform of motor load torque and stator current is more stable under no-load and load conditions， significantly improving the operational performance of the motor.关键词：electric excitation synchronous motor;co-simulation;maximum torque-current ratio;vector control;linear active disturbance rejection control;model assist - extended state observer
- 71
- |
- 13
- |
- 0
发布时间：2024-01-29 - 摘要：Permanent magnet synchronous motor （PMSM） is widely used in AC servo systems， and its mathematical model has the characteristics of nonlinearity， multivariability， and strong coupling. In order to improve the efficiency of modeling and analysis， while also considering the generality of PMSM models in complex multi-domain systems， a modeling method for AC servo-control systems based on Modelica is proposed， the non-causal modeling of PMSM is realized by using the way of modular modeling and Modelica language， and a simulation model of the servo system is established on the MWORKS platform， three typical operating conditions have been tested in terms of position control and speed control， with small tracking error and fast dynamic response. The correctness and rationality of the modeling method and system model construction are verified， which provides several new ideas for the future development of the servo-control system and multi-domain unified modeling and simulation technology.关键词：permanent magnet synchronous motor;servo-control;MWORKS;Modelica;multi-domain modeling
- 42
- |
- 6
- |
- 0
发布时间：2024-01-29 - 摘要：The sorting load of the string fruit sorting parallel robot is unknown and dynamically changing. When the fruit string is entangled， it will cause a significant change in the rotational inertia of the load. To achieve high-performance control of a string fruit sorting parallel robot， an adaptive sliding mode control algorithm is proposed that can identify the system load moment of inertia online. On the basis of analyzing the mechanical motion equations of AC servo motors， a gradient correction parameter identification algorithm is used to identify the load moment of inertia， and an adaptive rule is designed to improve the ability of parallel robot systems to overcome load changes while effectively suppressing chattering caused by sliding mode control. The control algorithm was simulated on MATLAB and applied to the prototype platform of a serial fruit sorting parallel robot for experiments. The experimental results verified the effectiveness of the algorithm in identifying load changes.关键词：parallel robot;online identification;moment of inertia;sliding mode control
- 33
- |
- 5
- |
- 0
发布时间：2024-01-29 - 摘要：In the parallel channel of the nuclear reactor core， flow instability can lead to a significant decrease in critical heat flux （CHF） or mechanical oscillation of the core fuel assembly， thereby endangering the normal operation of the reactor. In order to solve this problem， based on a high-temperature and high-pressure steam water two-phase thermal hydraulic experimental device， a parallel dual channel system model was built using NUMAP software for flow instability simulation research. The experiment shows that when the number of parallel channels is 2， 3， and 4， the flow instability boundary of the parallel multi-channel system under the same operating conditions is calculated through simulation， and the calculation error of the flow instability boundary is not more than ± 5%. It is verified that the flow instability boundary can be obtained by using a parallel dual channel structure when there are multiple parallel heating channels， and the transient analysis ability of the software for flow instability phenomena is confirmed.关键词：flow instability;parallel multi-channel system;flow instability boundary;Modelica;NUMAP
- 15
- |
- 5
- |
- 0
发布时间：2024-01-29 - 摘要：In order to make the electric servo loading system achieve better control effect， improve a series of problems such as poor torque following effect， large amplitude error and phase error， and difficult to apply complex control algorithm， an iterative learning controller based on fuzzy control is designed by using the characteristics of fuzzy control self-adaptation and high tracking accuracy of iterative learning.The algorithm is written in ST language， and the torque curve is collected by loading the experimental bench. It can be seen from the experiment that the iterative learning controller based on fuzzy control not only has the advantages of high tracking accuracy of ordinary iterative learning controller， but also has the characteristics of adaptive fuzzy control. It can achieve the required torque curve faster with less iterations， which proves the superiority and feasibility of the iterative learning controller based on fuzzy control.关键词：fuzzy control;iterative learning;electric servo loading;ST language
- 8
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：To improve the accuracy and real-time performance of cervical spine rehabilitation motion recognition， a cervical spine rehabilitation action recognition method based on joint point connection breadth matrix is proposed. In the preprocessing stage， the skeleton is first extracted from real-time human video stream data； Then， based on the action characteristics of the human skeleton， the feature information of the human skeleton during action is accurately extracted， and the change information in local space is obtained. The breadth first search algorithm is used to traverse the relevant nodes in the skeleton graph， establish the joint point connection breadth matrix， divide the triangular subgraph based on the connection breadth information， and assign weights to the joint points to improve the recognition efficiency of the model； Finally， the spatiotemporal features of the joint connection breadth matrix are extracted， and recognition is completed through SVM classifiers. The recognition method was validated on the CRED and MSR Action 3D public datasets for cervical rehabilitation movement. The experimental results showed that the average time consumption of the cervical rehabilitation movement recognition method based on the joint point connection breadth matrix was 1.20 seconds， the average frame rate was 27， and the average recognition accuracy was 92.72%， which has certain advantages compared to existing methods.关键词：cervical spine rehabilitation exercises;action joint connection breadth matrix;human skeleton;triangle subgraph;spatial temporal features
- 16
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：Unsupervised domain adaption （UDA） aims to transfer knowledge from the related and label-rich source domain to the label-scarce target domain. Usually， domain adaptation methods assume that the source data is correctly labeled. However， the labels and features of source samples will be destroyed due to the actual noise environment. To solve the problem of noisy source domain， this paper proposed noise correction domain adaptation based on classifiers discrepancy （NCDA）. First， this method made a more precise classification standard by the difference between multiple classifiers in the network， which can divide noisy source samples into feature noise samples， label noise samples， and clean samples. Second， different correction methods were applied on them. Then， the corrected samples were put back into the training procedure. Finally， this paper used the idea of stochastic classifiers to improve the network. Extensive experiments on Office-31， Office-Home and Bing-Caltech demonstrated the effectiveness and robustness of NCDA， whose accuracy is 0.2%~1.6% higher than the sub-optimal method.关键词：unsupervised domain adaptation;noise detection;noise correction;machine learning
- 16
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：A mixed integer programming model is established with the aim of minimizing the time which takes AMR to finally complete the picking task in the "goods-to-person" background mode. It is done by analyzing the picking mode and operation process of AMR and taking into account the speed of AMR itself， the number of pallets， and the load limit. A simulated annealing metropolis criterion that is simple to escape the local optimal is added to the standard particle swarm algorithm， and when combined with the dynamic inertia weight value and the penalty function of decreasing curve， a simulated annealing particle swarm mixing algorithm is created to solve the model. The model and algorithm are then validated using an intelligent picking center as an example. The simulated annealing particle swarm mixing algorithm is compared with the standard particle swarm algorithm and the simulated annealing algorithm， and the results show that the proposed model and algorithm outperform the other two heuristic algorithms in terms of iteration speed and solution deviation.关键词：goods to person;AMR;smart picking center;path planning;simulated annealed particle swarm mixing algorithm
- 6
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：A framework for abnormal behavior detection is proposed to address safety production issues in current industrial scenarios， mainly targeting two special situations： workers sleeping and falling. The idea of combining human key point recognition with machine learning classifiers is adopted. Firstly， key point recognition is performed on workers in video images， body coordinate point information is extracted， and then the classifier is trained for classification. Multiple machine learning methods and an integrated learning model are used to detect abnormal situations. On the fall dataset， the accuracy， accuracy， and recall of the ensemble learning algorithm reached 92.86%， 87.58%， and 98.96%， respectively； In terms of sleep detection， the accuracy， accuracy， and recall of the algorithm reached 98.51%， 95.81%， and 94.97%， respectively. Experiments have shown that this framework can effectively detect abnormal situations， help standardize production behavior， and has practical application value．关键词：behavior recognition;action recognition;abnormal behavior;fall detection;machine learning
- 20
- |
- 3
- |
- 0
发布时间：2024-01-29 - 摘要：To address the phase mismatch problem of single-channel speech enhancement techniques in the frequency domain， a joint time-domain and frequency-domain speech enhancement algorithm is proposed to jointly optimise the learning targets in different domains during the training phase. An attention mechanism is added to simulate human auditory perceptual characteristics to enhance the model's ability to suppress noisy signals. It also uses expanded convolution to widen the perceptual field， enabling the fusion of more input layer information and the effective extraction of local features in the time and frequency domains. To enhance the speech enhancement performance， the joint time-domain and frequency-domain loss functions are optimised for different domain learning. To validate the effectiveness of the proposed method， extensive experiments are conducted on the dataset VoiceBank using residual time convolution as the baseline model， and the experimental structure shows better enhancement than using a single baseline model in the time or frequency domain. The perceptual speech quality （PESQ） after denoising was 3.06 and the signal distortion ratio （SI-SDR） was 20.00.关键词：speech enhancement;time domain;frequency domain;TCN;attention module
- 18
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：Protein model quality assessment refers to the scoring of protein models predicted by computational methods， so as to select an excellent model that is closer to the native structure. Graph structures can intuitively represent protein models， so graph convolutional neural networks （GCNs） have been widely used in quality assessment in recent years. However， the fixed adjacency relationship of graph nodes limits the ability of GCN to mine node features deeply. Based on this， a dynamic graph convolution quality assessment method DGCQA is proposed to predict the global quality score of the protein model. This method dynamically obtains the neighborhood according to the feature distance of the node， and combines the multi-scale convolution module to extract the residue pair features to enhance the expressive ability of the network. In addition， based on the idea of transfer learning， the protein pre-training model ESM-1b encoding feature is introduced， which improves the performance of DGCQA on multiple indicators. The final experiments show that DGCQA is highly competitive in comparison with 12 quality assessment methods based on the CASP13 dataset.关键词：protein model quality assessment;dynamic graph convolution;transfer learning;ESM-1b
- 18
- |
- 2
- |
- 0
发布时间：2024-01-29 - 摘要：Currently， large-scale text question answering relies on sentence representation to retrieve answers from candidate texts， but it ignores that some answers require further reasoning and cannot be obtained directly from the text， such as judgment sentences. To solve such problems， a judgment sentence answer generation method for large-scale text is proposed.Firstly， in the semantic encoder， the semantic encoder is obtained by continuing to pre-train large-scale texts， and the questions and cues are semantically encoded. Sceondly，in the answer generator module， positive and negative samples are constructed based on contrastive learning for data enhancement. Then fast characterization and matching of questions and large-scale text is achieved by using Faiss in the answer basis obtainer. The accuracy of the final judgment sentence question and answer is as high as 96.58%， which verifies the effectiveness of this method.关键词：intelligent question answering;open domain Q&A;Q&A classifier;semantic coding
- 16
- |
- 2
- |
- 0
发布时间：2024-01-29 - 摘要：In virtualized environments， cloud servers consume a significant amount of CPU resources during data transmission between virtual machines， leading to decreased network performance. To address this issue， this paper proposes a kernel-based TCP/IP data processing optimization method （KOTOM）. This method aims to enhance the performance of cloud servers in virtualized environments through server kernel module software design. By establishing a Linux kernel cache and monitoring TCP/IP transmission content， KOTOM implements kernel-level caching for hot access data， thereby reducing the resource overhead caused by frequent switching between user and kernel modes during the parsing process of the TCP/IP protocol stack and improving network transmission efficiency. KOTOM utilizes a red-black tree to enhance cache lookup efficiency and adopts the LRU cache replacement strategy to implement cache replacement for hot data. Experiments show that using this method can reduce CPU utilization by up to 7% and increase data request response speed by 22%.关键词：virtualized environments;process optimization;netfilter;LRU;TCP/IP
- 14
- |
- 4
- |
- 0
发布时间：2024-01-29 - 摘要：With the rapid increase in data volume， distributed storage systems using regenerative codes as fault tolerance mechanisms need to use scaling techniques to expand their storage capacity. However， there are few existing methods for expanding the capacity of regenerative codes， and there is room for improvement in terms of expansion time and transmission volume. To this end， a scaling method S-E-MBR is proposed for online distributed storage scenarios， which reduces the number of migrated data blocks and I/O （Input/Output） overhead in a more efficient migration manner， reduces the required data blocks for verification updates， and achieves the optimal theoretical data block migration amount.Theoretical analysis and experimental results show that compared with RR and Scale-RS methods， the S-E-MBR method reduces data transmission by 52.7%~77.9% and 41.3%~50.4% respectively during expansion， reduces total expansion time by 72.3%~75.4% and 50.6%~53.5% respectively， and improves response speed by 39.2% and 17.1%， which can meet the needs of online expansion scenarios.关键词：regeneration codes;scaling method;distributed storage;erasure code;transfer volume
- 13
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：Link prediction algorithm based on node similarity usually predicts whether there is a link between node pairs according to the similarity between two nodes. A second-order link prediction method is proposed to determine if there is an unconnected node between node pairs， and then complete the second-order links between node pairs. The second-order link prediction index is used to calculate the similarity between known nodes and other nodes that do not have links， and the second-order reachable network is constructed to retain the second-order links in the original network. The experimental results show that the missing nodes between node pairs can be identified in real network data and their second-order links can be completed. The performance of different link prediction indices varies across four different networks， and the best precision rate reaches 83.7% in all experiments.关键词：complex network;second-order link prediction;reachable network;similarity index;common neighbor
- 10
- |
- 2
- |
- 0
发布时间：2024-01-29 - 摘要：In view of the dilemma of sensor data loss or missing due to the surrounding environment in traditional industries， a deep learning method based on causal analysis for multivariate data in energy systems is proposed in the case of unknown data distribution， and the missing value is supplemented by the results. First of all， rebalance the samples， and then a model is built based on LSTM's multivariate model. Causal analysis is used to optimize the deep learning optimizer and remove the influence factors that are not expected in the learning process. The pseudo-correlation between the eigenvalue and stable deflection is weakened， and the influence of stable deflection on the eigenvalue is excluded by placebo effect. Finally， the eigenvalue is subtracted from the harmful factor to obtain the value of removing the harmful factor， and then the model is optimized to obtain better results. This method solves the problem of underfitting the head data and overfitting the tail data in the process of machine learning. Experiments on multi-variable energy system data sets show that this method is more accurate in converging the missing value interpolation to the true value.关键词：causal analysis;neural network;long tail distribution;missing value interpolation
- 8
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：To use high-performance computing services， users need to go through a series of tedious steps such as multi end collaboration， manual configuration of information， and manual diagnosis of connection errors. To simplify the steps for users to use high-performance computing services， a one-stop service system for school level high-performance computing platforms has been designed and implemented. This service system integrates the functions required for users to log in and use high-performance computing clusters on a daily basis. It has a built-in terminal simulator， supports automatic information completion， and can automatically diagnose connection errors. In addition， the client of the service system has cross platform and highly configurable features， and supports modern interfaces. Testing and actual deployment have shown that the system can provide low latency services and still has high throughput under peak workloads of a large number of concurrent users， greatly improving the user experience of high-performance computing services.关键词：high performance computing;one-stop service;service system;cross platform;error diagnosis
- 13
- |
- 2
- |
- 0
发布时间：2024-01-29 - 摘要：In order to meet the requirements of multi-objective tasks in the performance test of the actuator， a performance test system of the actuator is designed. The system can complete the no-load performance test of actuator . With the cooperation of the four-channel loading platform， the load torque of the actuator in the air can be simulated to realize the loading test. The system based on LabWindows/CVI virtual instrument design platform can realize multi-point and multi-channel data transmission through RS422 transmission protocols， and generate a variety of test excitation signals. It can complete the functions of power-on test， signal feedback， data collection and information display of actuator. Through the test， the test results meet the requirements of the corresponding design， compared with the traditional actuator test system， improve the test efficiency and test accuracy.关键词：actuator test system;LabWindows/CVI;virtual instrument
- 7
- |
- 2
- |
- 0
发布时间：2024-01-29 -
HotellingT2 Small Sample Multivariate Process Quality Monitoring Method Based on Bayesian Theory AI导读
摘要：Traditional Hotelling T2 control chart method suitable for multivariate quality control has certain limitations in small batch processing due to its sensitivity to outliers due to limited sample data. To this end， Bayesian theory is combined with traditional Hotelling T2 control charts to estimate Bayesian parameters through historical batch process quality distribution information and existing real-time sample data. A Hotelling T2 process quality control chart based on Bayesian theory is constructed to resist the impact of outliers in the multivariate quality control process of small batch processing. The analysis of quality control examples for engine camshaft processes shows that the proposed method can effectively resist the impact of outliers compared to traditional methods， has good anti-interference and stability， and can better monitor the uncontrolled state of process quality control.关键词：multivariate quality control;Bayesian theory;HotellingT2 control chart- 6
- |
- 8
- |
- 0
发布时间：2024-01-29 - 摘要：The main task of small object detection is to detect images with dimensions smaller than 32×32 pixel target and classify it. Due to the inaccurate matching of traditional rectangular anchor frame structures in detecting small targets， the number of small targets in the general dataset is small and their distribution is uneven， which will lead to poor model detection performance. Therefore， based on Faster R-CNN， a small target detection method with circular anchor frames is proposed. In the RPN stage， a circular anchor frame is used to locate the region of interest， and a new area intersection and union ratio calculation method and loss function are used to reduce the model parameter quantity and offset calculation in the anchor frame regression stage， in order to enhance the model's fitting ability to the detected target and improve the model's detection accuracy and efficiency. At the same time， in order to address the issues of low proportion and uneven distribution of small targets in existing public datasets， data augmentation was performed on the MS COCO 2017 dataset， retaining only the small targets and modifying the annotation information to a circular bounding box with a high wrapping rate for the small targets. Experiments have shown that the circular anchor box method and data augmentation method have better detection performance in detecting small targets， with detection efficiency and speed significantly better than Faster R-CNN， APS and detection speed have been improved by 4.1% and 4 FPS， respectively.关键词：small target detection;Faster R-CNN;circular anchor;data augmentation;circle intersection over union
- 22
- |
- 1
- |
- 0
发布时间：2024-01-29 - 摘要：Aiming at the problems of missing detection and false detection of small target pedestrians with small scale and dense pedestrians， this paper proposes a pedestrian detection algorithm with mixed attention mechanism and C2f module based on YoloX algorithm. In this algorithm， firstly， the BAM module and the C2f module are fused to effectively enhance the characteristics of pedestrians， reduce the amount of calculation， and improve the detection speed. Secondly， the attention mechanism is used to guide the network to pay attention to pedestrian targets， and the characteristic information of pedestrian targets has been further strengthened. Finally， experimental analysis is carried out on the Crowd Human dataset， when the IoU threshold is set to 0.5， the small-scale pedestrian detection accuracy is 21.1%， the mesoscale pedestrian detection accuracy is 47.3%， the large-scale pedestrian detection accuracy is 64.7%， the total pedestrian target detection accuracy is 73.2%， and the detection speed is 24.3 frames per second. Experimental results show that the pedestrian detection algorithm in this paper effectively improves the detection accuracy and detection speed of pedestrian targets， and has good detection performance for pedestrian targets.关键词：pedestrian detection;feature enhancement;YoloX;deep learning;attention mechanisms
- 15
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：Current autonomous driving technology focuses on safety. With the development of autonomous driving technology， people's requirements for driving comfort will also continue to increase. A YOLOv5-based target detection improvement method is proposed for detecting small and medium-sized obstacles that affect driving comfort. In order to solve the problem that small and medium-sized obstacles affecting driving comfort are very similar to the background， the CA （Coordinate Attention） module is introduced ， which improves the ability to extra the salient features of the target while keeping the model lightweight and improving the attention to the key information； The CIoU loss function is replaced by the α-IoU loss function as the bounding box regression loss function， which improves the optimization space for different levels of targets； The new convolution module is designed to retain the original features while incorporating deeper feature information and reducing the number of parameters. The experimental results show that the improved method improves the mAP （mean average precision） from 87.8% to 89.9% compared to the original YOLOv5 with the reduction of the number of parameters and GFLOPs， and the FPS of single image detection on GPU reaches 70， which is better than the comparison algorithm and improves the detection effect while satisfying the real-time performance.关键词：deep learning;object detection;YOLOv5;coordinate attention mechanism;α-IoU
- 18
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：Aiming at the problem that it is difficult for teachers to timely learn students’ engagement due to the separation of time and space in the online education environment， a lightweight deep learning network model is designed for the detection of students' engagement. The model makes decisions based on the student's facial expression information. It uses a deep residual network to extract spatial features and a long short-term memory network to extract temporal features. The Shuffle Attention and the Global Attention are added to optimize the feature extraction ability of the model to improve the effect of engagement detection. The experimental results show that the proposed method achieves high accuracy on both public and self-collected datasets. It is better suited to the needs of practical online learning scenarios in terms of accuracy and time cost.关键词：online education;engagement detection;ResNet18;LSTM;attention mechanism
- 7
- |
- 2
- |
- 0
发布时间：2024-01-29 - 摘要：Image compressed sensing （CS） reconstruction method aims to restore the sampled image to a high-quality image. At present， CS reconstruction algorithm based on deep learning has superior performance in reconstruction quality and speed， but it has the problem of poor image reconstruction quality at low sampling rate. Therefore， an image CS reconstruction network based on multi-scale attention fusion is proposed. Multiple multi-scale residual blocks are introduced into the network to extract the information of different sizes of images， and the spatial attention of each multi-scale residual block and the channel attention of dense residual blocks are fused. The local features and global dependencies are adaptively integrated to improve the quality of image reconstruction. Experimental results show that the proposed algorithm is superior to other classical methods in PSNR and SSIM， and has better reconstruction performance.关键词：compression sensing;attention mechanism;deep learning;multi-scale feature extraction
- 10
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：Body size parameter is an important indicator for evaluating the growth status of fattening pigs. It addresses the problems of single measurement parameters and large measurement errors caused by factors such as angle and light source in pig body size measurement using a monocular CCD camera. Firstly， use the depth camera KinectV2 to synchronously obtain local point cloud data of the pig body from the top and left and right perspectives； Then， point cloud denoising， simplification and segmentation are carried out， and the improved ICP point cloud registration technology is used to process point cloud information； Finally， precise estimation techniques are used to streamline point cloud data. Comparing the results of experimental and manual measurements from different angles， it was found that the average relative error of body length was 2.65%， the average relative error of body height was 1.87%， the average relative error of body width was 1.75%， the average relative error of hip height was 2.07%， and the average relative error of hip width was 1.96% in pig body data. Overall， the error was relatively small， proving the effectiveness of the proposed method and providing a new solution for pig body size measurement.关键词：growing and fattening pigs;image processing;KinectV2;3D point cloud;body size measurement
- 9
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：In the UAV aerial image target detection task， the traditional target detection algorithm is poor in real-time and accuracy. The original YOLO algorithm has a high error detection and omission rate for small targets. The requirements of aerial image are higher in view angle， image data amount， target scale and so on， which are significantly different from ordinary images. Therefore， an improved algorithm based on YOLO-v7， FCL-YOLO-v7， is proposed to solve the problem of small target detection in UAV aerial images. First， add small target detection layer， improve the feature extraction network structure and prior frame configuration； Secondly， the SiLU activation function is replaced by FReLU activation function. Thirdly， CBAM attention mechanism is added to the backbone network； Finally， the small target data set is constructed by combining the open data set and the autonomous UAV aerial images. The experimental results show that the accuracy of the improved algorithm is 6.7% higher than that of the original algorithm and 7.3% higher than that of YOLO-v3. The recall rate is 3.3% higher than the YOLO-v5.关键词：UAV aerial image;small target detection;feature extraction;activation function;attention mechanism
- 33
- |
- 3
- |
- 0
发布时间：2024-01-29 - 摘要：Depth semantic segmentation is one of the common remote sensing image applications. The existing semantic segmentation algorithms based on depth convolution neural networks can not be effectively applied to image segmentation tasks in real environments. Such network models have many parameters， complex calculation and slow operation. For this reason， this paper proposes an image segmentation network based on convolutional neural network and multilayer perceptron （MLP）， which includes a convolution stage and a MLP stage. An attention control mechanism is added in the process of the jump connection between the encoder and the decoder， so that the network will place more weight in places worthy of attention. The shift based MLP network proposed in this paper can effectively extract local features of images. At the same time， compared with other complex neural network models， the proposed method can effectively reduce the number of parameters and computational complexity， while maintaining the accuracy of segmentation. Finally， the method in this paper is tested on several remote sensing data sets. The results show that the parameters of the model in this paper are 1.471 93M， the average training time is 47.973 218 55s， and the computational complexity is 5.7 GFLOPs， compared with the UNet， UNet++， and SegNet models， which reduces the complexity and running time of the model to a certain extent.关键词：remote sensing images;semantic segmentation;lightweight;multilayer perceptron;deep learning;attention mechanism
- 12
- |
- 1
- |
- 0
发布时间：2024-01-29 - 摘要：Movie rating prediction aims to predict the possible ratings that users may give to unreviewed movies， and is an important basis for practical applications such as recommendation systems and movie classification. Existing prediction methods mainly focus on the representation of interaction information and text information between users and movies， with less consideration given to the direct representation of attribute features. To this end， a movie rating prediction model based on interactive attribute enhancement is proposed. Firstly， consider using the embedding vectors of attribute nodes in the network to represent different attribute feature information. Construct a movie information network based on the interaction and dependency relationships between data， and use the Metapath2vec algorithm to obtain the embedding vectors of attribute nodes. Convert each attribute feature into vector representations with different meta path structure information and semantic information. Then， the attribute feature vectors of users and movies are inputted into the two-tower model and interactively fused with their respective ID feature vectors to explore the impact of different attribute preferences on users and movies. Finally， the user and movie feature vectors are obtained， and the user's rating prediction for the movie is achieved through dot product. The results on public datasets indicate that the proposed model has higher prediction accuracy compared to traditional models， demonstrating the effectiveness of the model.关键词：movie rating prediction;Metapath2vec;two-tower model;interactive attribute
- 8
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：Sentiment analysis is one of the important directions in the field of natural language processing. Existing researches on the influence of context of exploration still has insufficient challenges such as difficulty in syntax information capture， loss of semantic information， and lack of semantic context. Aiming at these problems， propose a novel combination of local global context guidance network to improve the performance and expression ability in the aspect-based sentiment analysis. Specifically， in this method， a dependency syntax parsing tree is constructed firstly to introduce more diversified information features for the model； Then， by introducing the context focusing mechanism， the characteristics of the original text and dependency syntax parsing tree are refined. At the same time， the local feature vector of the refining is interacted with the global feature vector so as to retain the context feature information of the words effectively. Finally， the characteristic aggregation module is aggregated to the local global characteristics， which improves the accuracy of the model in emotional polarity prediction. The experimental results on the four benchmark datasets show that compared with the baseline models， the accuracy of the proposed model increases by 1.67%， 1.67%， 0.7% and 0.16% respectively， and the F1 value increases by 2.55%， 2.03%， 1.57% and 2.08% respectively.关键词：sentiment analysis;natural language processing;local context;dependency syntax parsing tree;information features
- 9
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：In order to overcome the problems of high altitude， poor traffic conditions and limited traditional exploration methods in Pulang ore district， hyperspectral remote sensing technology is used to extract rock and ore information more conveniently and accurately. Using Hyperion data， the pure pixel is extracted as the end element by the minimum noise separation transform and the pure pixel index method， and the end element spectrum is obtained by N-dimensional visualization analysis. Then， spectral Angle mapping is used to extract altered minerals in the study area， and combined with known geological data analysis， to verify the feasibility of this method. The results show that hyperspectral remote sensing can accurately identify the types of altered minerals and display the distribution range. At the same time， the results show that it can map the minerals and improve the availability of alteration information， which is of great significance for remote sensing extraction of alteration information in the Pulang copper ore district.关键词：hyper spectral;mineral mapping;Hyperion;mineral information extraction
- 6
- |
- 0
- |
- 0
发布时间：2024-01-29 - 摘要：The computer network is an important basic course for network engineering. To solve the problems in traditional course teaching and experimental teaching， taking the engineering education certification as the background， the teaching procedure of computer network is reformed based on outcome-based education（OBE）. Taking the course design framework as the point cut， the course reform and practice are conducted from the sides of the teaching goal， teaching procedure and teaching assessment. Firstly， the teaching goal of the computer network is optimized according to the graduation requirements， which clarifies the professional abilities required by the nation and society； Secondly， the drawbacks of the teaching procedure are complemented based on the framework of the teaching goal， which guarantees the implementation of teaching goal； Lastly， an education quality assessment system is designed based on the OBE concept， and the course improvement is continuously conducted based on the assessment results. The results of reform and practice show that course reform performs well， which lay a good foundation for the improvement of students’ professional quality.关键词：course teaching;experimental teaching;OBE;teaching goal;computer network
- 7
- |
- 2
- |
- 0
发布时间：2024-01-29 - 摘要：With the rapid development of information technology， the demand for talent in the IT industry is expanding year by year. In order to improve the quality of IT talent cultivation and promote the balance between supply and demand， this study collected 13 638 recruitment data from Zhaopin Recruitment to conduct research on the real talent demand of enterprises， excavate the characteristics of enterprise talent demand， and based on this， construct an iceberg competency model to construct an IT professional talent cultivation system. This study constructs an iceberg competency model from four dimensions： knowledge， skills， comprehensive literacy， awareness and attitude， providing reference opinions for talent cultivation and graduation employment guidance.关键词：personnel training of IT;iceberg competency model;text mining;characteristics of talent demand
- 23
- |
- 1
- |
- 0
发布时间：2024-01-29 - 摘要：Under the background of emerging engineering education and "Double first class" construction， a new demand for the training of engineering professionals is proposed. This paper takes "C language programming " for example to analyze pain points in the current course teaching. Meanwhile， combined with the characteristics of students' learning situation， taking students as the center and starting from the aspects of teaching concept， teaching means， assessment and evaluation， this paper explores the teaching reform ideas of theory， practice， ideology and politics， and designs a modular teaching practice of "one body and two wings" based on the knowledge graph reconstruction， which integrates the OBE idea and ideological and political elements to enhance students' practical ability， cultivate students' computational thinking， and highlight the innovation and value-leading of the course teaching. Ultimately， with the exploration and practice of this course， it will provide a reference for other programming courses to carry out teaching design.关键词：programming;first-class courses;teaching reform and practice;teaching mode
- 10
- |
- 2
- |
- 0
发布时间：2024-01-29
- 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