最新刊期
卷 22 ， 期 9 ， 2023
- 摘要：As a sub-task in the field of fine-grained sentiment analysis， aspect-based sentiment analysis （ABSA） aims to analyze the aspect-based sentiment polarity to a given target. The growing amount of feedback on social media has caught the attention of many researchers over the past decade. In order to adapt to ABSA in different scenarios， researchers continuously devise new techniques and approaches to handle diverse research questions. We first provide an overview of ABSA， and then describe the research methods and characteristics of ABSA in different scenarios for the fine-grained sentiment analysis subtask. Finally， by reviewing the emerging topics of ABSA， looking forward to the development trend of ABSA research， with a view to providing some reference for research in this field.关键词：aspect-based sentiment analysis;sentiment classification;deep learning;machine learning
- 125
- |
- 21
- |
- 0
发布时间：2023-09-15 - 摘要：3D reconstruction is an important aspect of computer vision. It has developed rapidly in recent years， and has been widely used in automatic driving， virtual reality and other fields. Among them， the image-based 3D reconstruction method is relatively mature， the algorithm process is clear. In this review， we do not only show the complete process from feature point matching， sparse point cloud reconstruction， dense point cloud reconstruction， 3D surface reconstruction and texture image creation， but also introduce typical technologies in depth， and make analysis and comparison. Finally， the the 3d reconstruction technology is summarized and prospected .关键词：image-based 3D reconstruction;sparse point cloud reconstruction;dense point cloud reconstruction;typical technology
- 75
- |
- 17
- |
- 0
发布时间：2023-09-15 - 摘要：Faced with the increasing demand of network users， the traditional TCP/IP architecture has long been overwhelmed， exposing drawbacks such as difficulty in ensuring service quality， low efficiency in routing fault recovery， and poor scalability. In order to meet these challenges and improve network disaster tolerance， scientists have introduced overlay network into the research of network disaster recovery. The overlay network can build a new logical network by reasonably selecting connection nodes on the basis of retaining the current Internet structure， so as to provide basic services such as routing， multicast， content distribution， etc. First， the development status of overlay network and the meaning of network disaster tolerance are introduced. Then， the relevant literature is sorted out and classified. Based on the application of overlay network in disaster recovery， the disaster damage fault modeling scheme and damaged network connection recovery strategy are discussed. At the same time， it points out the shortcomings of current research on disaster recovery based on overlay network， and looks forward to the future development direction.关键词：overlay network;network failure;disaster recovery
- 30
- |
- 6
- |
- 0
发布时间：2023-09-15 - 摘要：The virtual reality learning environment is a virtual environment including one or more teaching objectives based on a certain teaching mode， and learners build an immersive emotional experience in the virtual reality learning environment according to the virtual image and affective participation. With the popularization of VR learning environment， the research on affective experience in VR learning environment has grown rapidly， but most of them are empirical studies， lacking systematic sorting， which is not conducive to further research. Using the systematic literature review method to analyze 38 empirical research papers in the past ten years abroad， the affective experience factors in the VR learning environment are summarized as immersion， presence， embodiment and flow. Among them， immersion and presence are related to the learner's perception characteristics from the technical realization； Embodied feeling and flow are related to the learner's perceptual characteristics from the body and the situation. The measurement of emotional experience in the VR learning environment is mainly realized by the methods of physiological behavior observation and perceptual experience inquiry. Questionnaire is the most commonly used emotional experience measurement method. The findings of the systematic literature review can outline a research framework for the design and application of emotional experiences in virtual reality learning environments.关键词：virtual reality learning environment;affective experience;systematic literature review;immersion;presence;embodied feeling;flow
- 27
- |
- 4
- |
- 0
发布时间：2023-09-15 - 摘要：In the wave of global digitization， smart healthcare based on digital twin technology has gained widespread attention. By combining relevant literature analysis， research papers on smart healthcare using digital twin technology were retrieved from both the China National Knowledge Infrastructure （CNKI） and Web of Science （WoS） databases. CiteSpace was employed for visual analysis to explore the current status and trends of smart healthcare based on digital twin technology. Drawing from the application patterns of digital twin technology in fields such as aerospace， intelligent manufacturing， transportation， and western medicine， a four-tier technological framework for digital twin technology in traditional Chinese medicine diagnosis and treatment was designed and discussed. Additionally， an application approach based on a five-dimensional model for digital twin technology in traditional Chinese medicine diagnosis and treatment was explored. The aim of this study is to conduct an in-depth analysis and exploration of the application of digital twin technology in the field of traditional Chinese medicine diagnosis and treatment， with the goal of providing new ideas and directions for the innovation and inheritance of traditional Chinese medicine.关键词：digital twin;smart healthcare;TCM diagnosis and treatment
- 12
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：Real garbage dataset usually presents a serious long tail distribution phenomenon of unbalanced categories， which leads to the problem that the generalization of the traditional deep learning model is not high when performing waste sorting and recognition tasks. To this end， a new data re labeling algorithm and framework are proposed to improve the generalization and accuracy of cleaning robot recognition and garbage classification. This algorithm includes feature extraction， feature clustering， and label mapping modules. When training commonly used classification models， by analyzing the data distribution of the dataset， the feature vectors of the feature extraction module are input into the feature clustering module to generate several subcategories for each category， and corresponding pseudo labels are assigned to them to alleviate the problem of data imbalance at the label level. At the same time， during prediction， pseudo labels are converted into real labels through the label mapping module. The experiment shows that the proposed algorithm can significantly improve the performance of tail classes in garbage long tailed datasets without losing the performance of the head class， and the relabeling algorithm can significantly improve the classification accuracy of imbalanced learning methods for different categories in the baseline on long tailed garbage datasets.关键词：garbage classification;deep learning;class-imbalance learning;data relabeling;dataset analysis;feature clustering;image processing;computer vision
- 28
- |
- 14
- |
- 0
发布时间：2023-09-15 - 摘要：In order to solve the problem of insufficient information extraction and poor classification effect of a single deep learning model， this paper proposes a BA-InfoCNN-BiLSTM model with a hybrid multi-neural network. The model uses BERT as the word embedding layer to obtain the vector representation of the word， and then uses the attention mechanism to obtain different weights for the word， and then sends it to the improved text convolutional neural network on the one hand to obtain The local information features of the text， on the other hand， are sent to the bidirectional long short-term memory network to obtain the global information features of the text， and finally the extracted local information and global information. After the features are spliced and fused， they are sent to the softmax function for classification， and the classification results are obtained. After comparing experiments with other models， this model has achieved good classification results. It has achieved 95.07% and 84.95% accuracy on the Sina news data set and Sohu news data set respectively， which solves the problem of insufficient information captured by single model to a certain extent.关键词：text classification;word embedding;attention mechanism;convolutional neural network;recurrent neural network
- 22
- |
- 4
- |
- 0
发布时间：2023-09-15 - 摘要：The accurate forecasting of urban electricity load is extremely important to the operation and planning of power system， which can produce great economic value and social benefits. The medium- and long-term urban electricity demand forecasting needs to consider not only the long-term economic trend and seasonal cycle factors， but also many uncertainties and non-linear issues. However， the existing forecasting methods only consider part of these factors， making they can not achieve accurate forecasting. To fill this gap， this paper proposes a hybrid machine learning （HML） model for urban electricity load forecasting. Firstly， HML selects the important features that significantly affect the urban electricity load demands； Secondly， exponential smoothing （ETS） is employed to capture the seasonal and horizontal components of electricity load time series； Then， the long short term memory （LSTM） is used to discover the nonlinear trend of the time series of electricity load； Finally， ensemble learning is adopted to realize the effective aggregation of the performance of each learning module. In the experiments， the monthly electricity consumptions of two cities in China are adopted as the benchmark datasets. The results show that our HML model is superior to the latest existing models in terms of the forecasting accuracy of monthly electricity consumption.关键词：deep learning;exponential smoothing;long and short term memory;electricity load forecast;recursive neural networks;time series forecasting
- 24
- |
- 6
- |
- 0
发布时间：2023-09-15 - 摘要：Most existing deep learning models have a single structure， which usually reduces the ability to extract text semantic features. To this end， a sentiment analysis research model integrating dual channel semantic features （FDSF） is proposed. Firstly， the BERT pre trained language model is used to obtain the dynamic feature vector representation of the text. Then， the global semantic information extracted by the BiGRU Attention channel is adjusted by attention dynamic weights， and fused with the local semantic information extracted by the CNN channel for feature vectors. Finally， the fused features are processed through a fully connected layer and Softmax function to output the final emotional polarity. Experiment on online_shopping_10_cats of ChineseNLPcorpus， and dataset compiled by scholars Tan Songbo from the Chinese Academy of Sciences， compared with existing mainstream sentiment analysis methods， the FDSF model has the best F1 value and accuracy， proving its effectiveness and feasibility in sentiment analysis tasks.关键词：sentiment analysis;deep learning;bi-directional gated recurrent unit;semantic vector;dual channel
- 36
- |
- 2
- |
- 0
发布时间：2023-09-15 - 摘要：View-invariant is an essential factor in human action recognition. Aggregating human action information from different views can alleviate the effect of the obscured vision， and thus improve the accuracy of action recognition. In order to efficiently analyze various of video data from a limited number of views in real life and provide a highly available action recognition model， a self-supervised multi-view human action representation learning method based on contrastive learning （MAR-NET） is proposed， and it also provides a technical support for the actual deployment of edge intelligence. Specifically， input multiple short video clips from different views， and use multi-view data input to map the video to the feature embedding space， and then a new contrastive loss function is proposed to maximize the consistency of the same action feature representation in different views and constrains the distribution of different actions in the feature space. The features are compared for the end-to-end action self-supervised recognition task on actual edge devices. Through contrastive learning， the action recognition model proposed in this paper narrows the representation distance of two clips from different views from the same video in the embedding space， while distinguishing the representations between different actions， yet providing a high-performance solution for self-supervised action recognition on edge devices. In the experimental environment built by real edge devices， experiments demonstrate the effectiveness of MAR-NET in self-supervised multi-view action recognition. The method proposed in this paper improves the accuracy by 18.7% compared with the existing self-supervised action recognition method using only RGB modality.关键词：deep action recognition;edge computing;view-invariant;contrastive learning;end-to-end self-supervised learning on videos
- 14
- |
- 2
- |
- 0
发布时间：2023-09-15 - 摘要：With the growth of the asset volume， the traditional way of asset management by manual is challenged， so it is urgent to apply the asset management system to enterprises to achieve intelligent and efficient management. Most of the existing asset management systems focus on the realization of basic functions， and pay less attention to how to use visualization for assisting asset management better. Therefore， this paper proposes visual management that gives consideration to both visual effect and practicability. Through the visualization of RFID terminal intelligent read and write asset data， and realizes the interaction of scene visualization， business visualization， and data visualization. In addition， the visualization interaction mode is innovated from three aspects： interaction ability of data sets， configurable site layout， and humanized interaction scheme. The visual management system of station assets is designed and implemented， which has been successfully applied to Changning Natural Gas Company. The system can better serve the asset management through the visualization technology， so as to improve the efficiency of asset management， and has high promotion value.关键词：data visualization operation;station asset management;visual interaction;asset management system;RFID
- 40
- |
- 6
- |
- 0
发布时间：2023-09-15 - 摘要：In order to enable medical staff to understand the changes in cerebral blood oxygen levels after surgery in patients with clinical brain injury as soon as possible， and to make timely diagnosis and treatment， a non-invasive， real-time， and continuous monitoring device for cerebral blood oxygen levels is proposed. This device is based on near-infrared spectroscopy （NIRS） technology and has designed a dual wavelength light source （730 nm， 850 nm） and dual detector brain blood oxygen monitoring device； Select compact and portable SMD light emitting diodes and photodiodes as light sources and detectors， design hardware systems such as cross resistance amplification circuits， analog filtering circuits， and light source driver circuits； Write a concise GUI interface that utilizes the basic principles of cerebral blood oxygen monitoring to display real-time changes in cerebral blood oxygen on the upper computer. In the forearm occlusion experiment， tissue oxygen saturation showed a decreasing trend after occlusion， and gradually returned to resting levels after releasing pressure. In the experiment of measuring cerebral blood oxygen in different postures， the cerebral tissue blood oxygen saturation under reverse traction was significantly higher than that in resting sitting posture. The experimental results indicate that this device can monitor the changes in local brain tissue blood oxygen in real-time， preliminarily verifying the feasibility of the device.关键词：near infrared spectroscopy;cerebral blood oxygen;light source;detector;signal processing;forearm block
- 6
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：Potato is one of the most important staple crops in China. The loss caused by mechanical damage during the harvest period is an important factor affecting its high-quality harvest. At present， there have been many studies on the relevant damage mechanism， but the monitoring research directly applied to harvest scenes is rarely mentioned. To this end， a digital twinning system of potato tuber in harvest period was constructed by making a high simulation model of potato tuber and using sensor and digital twinning technology. Through real-time collection and transmission of the external force data borne by the potato tuber model in the real harvest process， the impact and collision borne by the potato are visually displayed in the form of 3D digital twins and data boards， and the skin damage warning is carried out. The research shows that the proposed method can be applied to adjust and assist the harvesting machinery， reduce the potato loss caused by various bumps and impacts， provide a solution for the damage monitoring and data collection of agricultural products in the harvest stage， and provide a feasible scheme for the application of digital twinning technology in the harvest stage of agricultural products.关键词：digital twin;potato harvest stage;damage;monitoring system;Unity3D
- 14
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：Cloud computing courses have the characteristics of multiple knowledge points， high difficulty， and uneven professional foundations for students， requiring them to invest more time in independent learning. Considering that there is currently no dedicated learning platform for cloud computing courses， and this platform not only needs to meet students' learning needs， but also needs to have the ability to run on systems such as Windows， Android， and iOS. To this end， based on the popularity of WeChat and the ability of WeChat mini programs to run across systems， a cloud computing learning platform has been developed. Five functional modules have been built for learning， answering questions， courseware management， textbook management， and statistical analysis， including login， answering questions， and review. Practice has shown that the platform can not only operate stably across systems， but also stimulate students' interest in learning by improving their learning experience， thereby significantly reducing the failure rate of courses.关键词：cloud computing;learning platform;WeChat mini program;cross system;learning interest
- 49
- |
- 5
- |
- 0
发布时间：2023-09-15 - 摘要：This paper takes the advantage of knowledge interconnection and visualization to construct a knowledge graph of the government’s emergency management of public health emergencies. It combed the emergency policies’ transmission and deployment measures of various management departments， established a multi-input coordination and linkage mechanism to improve the governance efficiency of public health emergencies. The policies on COVID-19 were taken as examples， a huge knowledge database of emergency management policy was constructed through the process of ontology modeling， BiLSTM+CNN-CRF algorithm for entity relationship extraction， knowledge reasoning based on rules and representations， and neo4j for graph’s visualization， etc. It's proved to have a great quality by sampling results of the graph.In this way a visual information transmission network is presented， which improves the status quo of the lack of relevance of common policies and disorderly information transmission in public health emergencies. It creates a fast emergency response channel for the formulation， release， and circulation of emergency management policies， and overcomes the shortcomings of emergency management in special periods.关键词：public health emergencies;knowledge graph;emergency management;visualization graph construction
- 39
- |
- 6
- |
- 0
发布时间：2023-09-15 - 摘要：Aiming at the problems of sparse data and low recommendation efficiency in traditional collaborative filtering recommendation algorithms， this paper proposes a collaborative filtering recommendation algorithm based on fuzzy clustering and user interest. Firstly， the user-item rating matrix and user-interest preference matrix are constructed respectively. Secondly， the fuzzy clustering algorithm based on particle swarm optimization is used to cluster user interests， and the average score of the item by users with similar preferences is filled into the user-item scoring matrix， which alleviates the sparsity of user data. Finally， the algorithm comprehensively considers the user-item rating matrix and user-interest preference matrix to calculate user similarity， and introduces the item type penalty factor to further improve the accuracy of user preference. Experimental results show that the proposed algorithm can effectively alleviate the sparsity of data and improve the accuracy of the recommendation algorithm.关键词：collaborative filtering;fuzzy clustering;particle swarm optimization algorithm;interest preference
- 17
- |
- 6
- |
- 0
发布时间：2023-09-15 - 摘要：Aiming at the problem that the current alliance chain consensus algorithm cannot achieve low latency， high throughput， and high security at the same time， a fault-tolerant Raft consensus algorithm suitable for alliance chain-BRaft （PBFT-Raft） is proposed. BRaft used digital signature to solve the problem of Byzantine leader node tampering with logs， and based on the three-stage agreement of PBFT algorithm， it introduced the flag W， which solve the problem of Byzantine follower node maliciously responding to leader node， and ensure that the error message sent by Byzantine node can be prevented. In this case， the log entry can still be submitted correctly. The experimental results show that BRaft improves the security of the algorithm while ensuring the comprehensibility and consensus efficiency of the algorithm.关键词：Raft algorithm;PBFT algorithm;Byzantine node;digital signature;identification bit
- 16
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：Video streaming application scenarios are widespread and common， but there are problems with low transmission quality and poor video quality. To this end， a segmented routing multi-path QUIC transmission framework （SDNMQS） for video streams was designed using the programmable characteristics of software defined network （SDN） and fast UDP Internet connection （QUIC）. This framework can dynamically allocate routes based on network conditions， achieving high-quality transmission services. The simulation test results on the simulation platform show that SDNMQS outperforms other traditional methods in various performance tests and can improve the quality of video transmission， providing reference for optimizing video transmission..关键词：QUIC;software defined network;segmented routing;multipath transmission;video streaming
- 12
- |
- 3
- |
- 0
发布时间：2023-09-15 - 摘要：Advanced persistent threat （APT） attacks on important facilities such as information systems in various countries are becoming increasingly frequent， and APT has the characteristics of strong targeting， good concealment， and high destructive power. In order to efficiently detect APT attacks， a knowledge-based APT attack detection scheme is proposed. Firstly， by collecting a large amount of open-source APT threat data， a novel APT knowledge acquisition method based on deep learning cascading model structure is proposed. Then， aiming at the multi-source heterogeneity of data， a semi supervised bootstrap knowledge fusion method is proposed to automatically build the APT Knowledge graph. Next， in order to improve the accuracy of APT attack detection and recognition， an APT attack detection scheme based on the Bert+BiLSTM+Self Attention+CRF model is proposed. The Bert model extracts text features， BILSTM extracts the relationship between input statements and context， and integrates the Self Attention mechanism to focus on the semantics and relationships between APT entities mentioned above and below. The CRF model extracts the globally optimal output label sequence based on the dependency relationship between labels， naming the entity to obtain APT attacks. The experiment shows that the accuracy， recall， and F1 values of the Bert+BiLSTM+Self Attention+CRF model reach 88.69%， 77.13%， and 82.5%， respectively， indicating better overall performance compared to existing methods.关键词：knowledge graph;advanced persistent threat;deep learning;APT attack
- 6
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：Cyber-physical system is a networked physical device that integrates computing， communication and control functions on the basis of environment perception. Through the feedback loop of mutual influence of computing process and physical process， the dynamic interaction and cooperation between the real world and the virtual world are constantly completed. At present， there are three kinds of network attacks in the field of Cyber-physical system research. Denial of service attacks exploit the weakness of network transmission protocol， consume a lot of system resources， and seriously threaten network security. Based on distributed network system， an anomaly detection model based on optimal state estimation and malicious attack detection is proposed by combining Kalman filter and multi-sensor fusion algorithm to solve the problems of natural measurement data loss and malicious interference attacks in unreliable wireless transmission channels. Finally， the effectiveness of the proposed fusion structure and detection model is verified by numerical simulation with MATLAB.关键词：cyber-physical system;Kalman filter;distributed;denial of service attack;KL detection
- 9
- |
- 2
- |
- 0
发布时间：2023-09-15 - 摘要：In order to solve the problem of large calculation of 3D reconstruction and large error of 3D model caused by the large number of video frames and easy to appear fuzzy frames， an adaptive step video key frame extraction algorithm is proposed. The algorithm calculates the image similarity based on the multi-channel histogram Euclidean distance， and the image sharpness based on the Laplacian gradient function. The key frame extraction step is dynamically determined according to the similarity between video frames， and the fuzzy frame is avoided through the sharpness detection. Experiments are carried out on high-definition constant speed video and variable speed blurred video. Compared with the current multiple video key frame extraction algorithms， the key frame extraction time of the proposed algorithm is reduced significantly， and the accuracy of the reconstructed 3D model is better.关键词：3D reconstruction;key frames;adaptive step size;image similarity;image sharpness
- 16
- |
- 5
- |
- 0
发布时间：2023-09-15 - 摘要：Landsat-8 OIL images of the Erhai Sea in Dali， Yunnan Province， are used as the base data. The Canny operator is combined with the water body index method commonly used in remote sensing to automatically extract the vector data of the lake boundary， and then the extraction accuracy of each water body index method is compared by visual discrimination of the high-resolution images and spatial buffer analysis. To address the situation that the traditional Canny edge detection algorithm requires artificially set thresholds， the maximum threshold segmentation method with interclass variance is used to optimise the high threshold setting principle in the Canny operator to ensure the adaptiveness of high and low thresholds after analysing the grey-scale histogram of the water body image with an obvious bimodal distribution. The results show that the application of the interclass variance maximum threshold segmentation method to the threshold selection of the Canny operator effectively improves the signal-to-noise ratio， and the lake boundaries extracted based on remote sensing imagery using the normalised difference water index （NDWI）， the improved NDWI， and the enhanced water index （EWI） combined with the Canny operator all achieved satisfactory results， with the improved NDWI achieving better extraction accuracy compared to the other two indices.关键词：multi-source remote sensing image;Canny operator;Otsu method;extraction method of water boundary
- 20
- |
- 5
- |
- 0
发布时间：2023-09-15 -
Research on Lightweight Self-Attention Mechanism as Backbone for Natural Land Cover Segmentation AI导读
摘要：Taking the 16-level satellite remote sensing images in the central and western regions of Zhidan County， Yan'an City， Shaanxi Province as the research object， a semantic segmentation method GFormer， which acts on low-resolution remote sensing satellite images of land and vegetation， is proposed. GFormer uses MixTransFormer as the backbone encoder and uses a newly designed decoder with progressive feature fusion structure. On this dataset， GFormer shows strong segmentation stability and segmentation ability. Compared with DeepLabV3+ and Unet with convolution as the backbone， GFormer shows the strong robustness of the semantic segmentation algorithm with attention mechanism as the backbone； compared with SegFormer which also uses MixTransFormer as the backbone encoder ， GFormer's newly designed full progressive fusion structure encoder has stronger decoding ability in remote sensing image segmentation. Compared with the model B5 with the best phenotype on SegFormer， MIoU and Kappa are increased by 3.25% and 3.04% respectively when the model scale is smaller； compared with SegFormer-B4 using the same scale MixTransFormer encoder， GFormer benefits The MIoU and Kappa of the newly designed progressive decoder are increased by 5.01% and 4.81% respectively. The experimental results demonstrate the effectiveness and robustness of GFormer in the segmentation of ground cover remote sensing images.关键词：computer vision;deep learning;semantic segmentation;self-attention mechanism- 11
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：Facial expression recognition is one of issue in intelligent human-computer interaction research. Facial emotion changes are closely related to areas of interest such as the mouth， eyes， eyebrows， nose， etc. These features are very important for recognizing facial expressions. To this end， an attention Gabor convolutional network consisting of four Gabor filtering convolutional layers， an attention module， and two fully linked layers is proposed， and the network is optimized using imbalanced loss focal loss. Firstly， Gabor directional filters modulated by Gabor kernels and traditional convolutional filters can better capture information about regions of interest compared to traditional convolutional filters. Then， channel attention and spatial attention modules are used to extract more critical features in the region. The experiments on FERPlus and RAF-DB datasets show that the model has a simple structure， is easy to train， and has low computational costs. The recognition accuracy reaches 88.39% and 87.22%， respectively.关键词：facial expression recognition;Gabor orientation filter;Gabor convolutional network;spatial attention module;channel attention module
- 17
- |
- 3
- |
- 0
发布时间：2023-09-15 - 摘要：Classification and detection of fruit quality is significant for fruit producing， processing， transporting and selling. In order to improve the precision and speed of detection， propose a fruit quality detection and classification scheme which using the detection algorithm YOLOv5 based on fruit images. Firstly， four kinds of fruit images including apples， oranges， bananas and pears are collected， and fruit images are labeled by the LabelImg tool to construct a fruit image dataset. Secondly， the YOLOv5s training method is given to obtain a fruit quality detection and classification network model with excellent performance. The results show that the average detection accuracy of YOLOv5s for fruit quality detection and classification can reach 95.3%， which is 3.7%， 0.2%， 13.1% and 8.73% compared with YOLOv3， YOLOv3-spp， YOLOv3-tiny and YOLOv4-tiny， respectively. The inference time of the algorithm per image is 10.5 ms， which has a certain degree of advancement. The use of YOLOv5 algorithm to detect fruit can provide new ideas for fruit quality detection in actual agricultural scenarios.关键词：fruit quality detection;deep learning;YOLOv5;object detection;fruit dataset
- 28
- |
- 5
- |
- 0
发布时间：2023-09-15 - 摘要：Cross domain clothing retrieval is a challenging task due to the large differences between domains， making it difficult to accurately retrieve. The existing cross domain garment retrieval algorithms based on convolutional neural network lack the use of local garment feature information， resulting in poor performance. A cross domain clothing image retrieval method combining feature fusion is proposed to address this issue. Based on deep convolutional neural network extraction， this method uses multi-scale convolution and self attention to extract representative local information， uses Gem pooling to extract global information， and aggregates local information with global representation to generate feature embedding more suitable for cross domain image retrieval. At the same time， the training process is constrained by the Loss function of ternary loss， center loss， classification loss and centroid loss， and the centroid loss is used in the retrieval phase to shorten the retrieval time. This method achieved good retrieval performance in the DeepFashion dataset， with a top-50 retrieval accuracy of 0.864， which is 1.4% higher than the CTL method. The cross domain clothing retrieval method that integrates global and local features can effectively improve retrieval accuracy while ensuring high retrieval efficiency.关键词：clothing retrieval;feature fusion;cross-scene;centroid loss
- 9
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：The spatial structure characteristics of Great Buddha Temple of Zhangye were studied by modeling through the spatial syntax Depthmap software and quantifying the connection value， integration value， depth value， agglomeration coefficient and comprehensibility of the Big Buddha Temple space from the visible layer and feasible layer. The results show that： ①the plants in the entrance area of Zhangye Big Buddha Temple block the penetration of the sight line and increase the mystery and solemnity of the temple； ②the distribution of the overall space of Zhangye Big Buddha Temple is dispersed， and the sight line radiates from the central space to the surrounding；③the roads and gardens divide the space and play a guiding role to the visitors and increase the amusement. The comprehensibility value of the space of Great Buddha Temple of Zhangye is high， and the spatial qualities are dominated by human factors， lacking the mysterious spatial characteristics of the temple and not reflecting the organic integration of the temple and the natural environment. The above analysis results can provide a reference basis for the future construction and repair of Great Buddha Temple of Zhangye.关键词：Great Buddha Temple of Zhangye;space syntax;spatial characteristics;quantitative analysis
- 4
- |
- 1
- |
- 0
发布时间：2023-09-15 - 摘要：Aiming at the problems of unclear definition of innovation，creativity and entrepreneurship in software engineering majors at home and abroad， and weak practice links， this paper sorts out the requirement index system of software engineering innovation and entrepreneurship ability under the background of new engineering. The exploration and practice process of curriculum reform is elaborated in terms of construction ideas and curriculum practice，and finally the practical results of the above-mentioned reform measures are introduced.关键词：new engineering;software engineering;innovation， creativity and entrepreneurship;curriculum construction;teaching reform
- 18
- |
- 5
- |
- 0
发布时间：2023-09-15 - 摘要：In order to meet the urgent need for the construction of composite and innovative talents in the process of digital transformation and upgrading of automobile industry under the background of " software-defined vehicle"， analyzed the employment situation of automobile industry and the talents demand of some automobile enterprises， proposed a software engineering talents cultivation model for automobile industry. This work also analyzed and discussed curriculums optimization， faculty development， and collaborative education mechanism. The model proposed in this work is to cultivate composite and innovative talents so that to meet the talents demand of automotive industry， thereby to meet the development of automobile industry and serve the national intelligent automobile innovation development strategy.关键词：emerging engineering education;automotive industry;software-defined vehicle;software engineering;compound innovation
- 9
- |
- 2
- |
- 0
发布时间：2023-09-15 -
Teaching Reform Research and Practice of GIS Secondary Development in the Era of Spatial Big Data AI导读
摘要：At present， spatial data collection methods are rapidly developing internationally. In China， multiple geographic information projects of national level have been implemented on a large scale， resulting in a huge amount of spatial data. It is now urgent to use GIS secondary development technology for batch and rapid spatial data mining. Traditional GIS secondary development technology was difficult， inefficient， and not competent for big data analysis， which led to the urgent need for reform in GIS secondary development teaching. This paper analyzed the transformation of GIS secondary development mode and secondary development language under the background of spatial big data， discussed the development environment， and designed the teaching content. Three countermeasures against the pain points of GIS secondary development， including automatic code generation， exploratory program development， and template framework code segment， was put forward. Finally， two actual positive and negative project teaching cases was comparatively analyzed. The teaching reform of GIS secondary development with geographical big data analysis as the core could significantly improve the learning effect of surveying and mapping and geographic information science majors， and increase the scientific and technological content for actual project research and development， hence achieve greater scientific research benefits， economic benefits and social benefits， and improve the construction capacity of "first-class majors".关键词：GIS secondary development;geographic information science;big data;project teaching;programming- 13
- |
- 2
- |
- 0
发布时间：2023-09-15 - 摘要：According to the challenges faced by international students in the field of computer science， such as language barriers and limited programming skills， this paper proposes a reform approach for the interactive experimental component of the compiler principles based on ChatGPT. Considering ChatGPT's cross-lingual， content generation， and context-aware capabilities， different prompt templates can be made for international students to obtain the personalized solution from the large-scale pretrained language model， to overcome language barriers and enhance their understanding and application of compiler principles knowledge. The incorporation of ChatGPT in the programming practice has a positive impact on improving the learning outcomes and programming skills of international students.关键词：compiler principle;ChatGPT;international students;programming practice;Prompt
- 24
- |
- 3
- |
- 0
发布时间：2023-09-15 - 摘要：In view of the prominent problems existing in the current teaching contents and teaching methods of data ＆ computer communications network course，the effective course construction methods are explored based on the objects of “high-level property， innovative property and challenge degree” （referred to as “two peoperties and one degree”）. Taking students as the teaching center， the course-learning goals of integration of knowledge， ability and quality are formulated， and the breadth and depth of the contents are expanded to improve the high-level nature of the course. An online and offline hybrid teaching mode are applied using the intelligent learning space online-platform， and a variety of teaching activities are developed to promote students to carry out the personalized learning. A diversified assessment and evaluation method that combine the qualitative and quantitative assessment， process and final assessment is designed， which increases the challenge of the course. The course construction oriented to “two peoperties and one degree” has achieved preliminary results and has been approved by most students.关键词：two peoperties and one degree;data and computer communication networks;hybrid teaching;course evaluation
- 4
- |
- 2
- |
- 0
发布时间：2023-09-15 - 摘要：In order to understand the learning situation of the students in the course of Introduction to Computer in a timely manner， the four-stage analysis method of learning behavior is proposed by correctly guiding the learning methods and adjusting the teaching strategies in a timely manner. Relying on the Learning Intelligence Education Platform to obtain learning data， normalize the learning behavior data of different dimensions to four intervals， and excavate the laws contained therein in the form of bubble chart， analyze the causes of behavior， carry out targeted intervention， and intervene and guide the learning behavior in the later stage of the course according to the learning behavior portrait in the early stage of the course. The practice shows that this method has significantly improved students' learning indicators， and the excellent ratio and passing rate of the final examination have increased by 6.19% and 10% year on year， which proves that this method can reveal the rules of learning behavior in mixed teaching， effectively guide students' learning behavior， improve learning efficiency and teaching effect， and also has reference significance for courses such as Introduction to Computational Thinking and Computer Culture Foundation.关键词：learning behavior;behavior portrait;learning portrait;analysis of learning situation;behavior intervention;data visualization;four-stage analysis
- 9
- |
- 2
- |
- 0
发布时间：2023-09-15 - 摘要：With the development of AI technology， the education form has been upgraded from "Internet+" education to "AI+" education. Artificial intelligence has a wide range of application scenarios in education， including personalized learning， adaptive evaluation， intelligent tutoring， and virtual reality teaching. However， the application of these technologies has also triggered a series of ethical crises， such as weakening the status of teachers， infringing on student freedom， and deepening educational inequality. By examining the ethical dilemma in the application of artificial intelligence education， including the unfairness of the algorithm shackles and the digital divide， the hidden danger of data ownership， and the emotional crisis of human-computer interaction， the root causes of the ethical dilemma are deeply analyzed， including the arrogance of instrumental and value rationality， the breach of professional literacy， and the lack of ethical conventions and supervision. To prevent and avoid these ethical issues， propose the path of "people-oriented principle requirements""value orientation of technology for the good"， and "ethical regulations guaranteed by law" to help artificial intelligence technology be more ethically applied in the field of education.关键词：artificial intelligence;educational applications;ethical dilemma;avoidance path
- 9
- |
- 3
- |
- 0
发布时间：2023-09-15 - 摘要：With the development of economic and social globalization， effective cooperation and collaborative problem-solving among people in different fields are particularly important. Collaborative problem-solving ability has become one of the essential key skills for citizens in the 21st century. At present， the main research on collaborative problem solving is usually the ability evaluation， cultivation and influencing factors. The research on how to cultivate college students' collaborative problem solving ability in the mixed learning environment is relatively rare. Therefore， combining the advantages of classroom learning and online learning， designed a cooperative problem solving activity model in a mixed learning environment， carried out practical research relying on modern educational technology courses， improved and revised the teaching model， and explored ways to cultivate college students' cooperative problem solving ability in the mixed learning environment.关键词：collaborative problem solving;blended learning;teaching design;online teaching;curriculum design
- 8
- |
- 3
- |
- 0
发布时间：2023-09-15
- 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