最新刊期
卷 21 ， 期 3 ， 2022
- 摘要：To encourage student participation in open source software by submitting high quality code and allow flexibility in selecting the open source software in which students are interested in, proposed software engineering education based on open source software. Provide detailed instructions and taught several automated program analysis tools to enable large scale integration of open source software into software engineering education. The proposed software engineering project have shown promising results in encouraging student contribution to open source software. The evaluation results show that most students like the proposed project and will recommend it as future Software Engineering course project. Moreover, students have also contributed many patches to open source software.关键词：software engineering education;open source software;lesson plan;course project
- 195
- |
- 53
- |
- 0
发布时间：2022-03-28 - 摘要：Considering the requirements and the problems of training system development ability in programming practice course, we propose the course task of developing an interpreter for a simple MIPS assembly language and a three-stage course organization. Besides, we have carried out several reforms and developed an automatic testing platform for the course. These reforms achieve a good teaching result.关键词：system development ability training;programming practice;comprehensive practice;interpreter
- 126
- |
- 143
- |
- 0
发布时间：2022-03-28 - 摘要：Software Architecture is the core course of software engineering major. Due to the characteristics of the course knowledge, in the teaching of this course, it is easy to appear profound and abstract knowledge, vague and unrealistic teaching content and other problems, which leads to the inevitable weakness of students in the application of architecture knowledge to enhance their engineering ability of software architecture design, implementation and application. Facing the problem, proposes a teaching objective, which is combination of learning and application, to be a sentient and warm software practitioner. In addition, carry out teaching exploration and practice in the aspects of characteristic teaching methods, course organization and feedback mechanism, in order to eliminate the huge gap between knowledge learning and engineering application, build a bridge between "learning and application", and achieve the cultivation goal of moral and intellectual two-way development of software engineering talents.关键词：software architecture;educational reform;mutual learning and application;software engineering;talent cultivation
- 48
- |
- 24
- |
- 0
发布时间：2022-03-28 - 摘要：With wide applications of modern software development practices, such as agile methodology, continuous integration, DevOps, in software industry, the requirement from software companies on professional software developers is changing. It is an important educational research question on the adaption to the requirement changes in professional degrees education. Education on engineering capabilities for professional post-graduates is proposed on the perspective of software process management. Aiming at software engineering capabilities, this course emphasizes the combination of the theory of agile methodology and the agile practices in modern software companies and facilitates the connection between professional graduate education and the industry, with the help of practice education on software engineering capabilities such as requirements management, incremental iterative development, code quality management and automatic tests, continuous integration and continuous delivery, enabled by enterprise-level cloud development platform. Education practices indicate that developing software process courses for graduates based on cloud development platform is beneficial for the students to improve their software engineering capability and for matching the requirements of professional developers in software companies.关键词：professional degree;software engineering capability;software process;practice-oriented teaching;continuous integration
- 43
- |
- 11
- |
- 0
发布时间：2022-03-28 - 摘要：Graduation project plays an important role in the training program of various majors in universities, which helps students get a full and thorough understanding of the professional knowledge and skills learned during the undergraduate period. To ensure the graduates achieve the standards of the Washington Agreement, according to the existing problems and the requirements of engineering education accreditation for graduates majored in software engineering, the course objectives of graduation project were redesigned, the teaching process was reorganized, new evaluation mechanism was put forward. The teaching quality of graduation design was quantitatively analyzedbycalculating the achieving degree. Targeted improvement measures are put forward to form a closed loop feedback of the teaching process. Finally, the effectiveness of this reform is illustrated by the actual data of 2020 and 2021.关键词：higher engineering education accreditation;software engineering;graduation design;achieving degree of course objectives
- 36
- |
- 18
- |
- 0
发布时间：2022-03-28 - 摘要：Aiming at the shortcomings of outdated course content and heavy dependence on foreign technologies in the software engineering student training system, through the analysis of the whole process of the first HarmonyOS application development course with the cooperation between Wuhan University and Huawei company, which based on the project of the Ministry of Education, this paper shows that the introduction of HarmonyOS technology has reformed the content of mobile application development course in Wuhan University. The reform enhances students'confidence in learning domestic software. The mode of college-enterprise cooperation courses can promote the reform of teachers'teaching content. At the same time, cooperating with domestic hardware and software companies can cultivate students' "four self-confidence" and a reserve force for the development of independent hardware and software.关键词：college-enterprise cooperation;co-built Hongmeng course;domestic hardware and software;HarmonyOS application development;softwere engineering
- 59
- |
- 26
- |
- 0
发布时间：2022-03-28 - 摘要：In order to improve the problems, such as hard implementation and weak effectiveness, during ideological and political education in the core curriculum of science and technology in colleges and universities. In the background of new engineering and technical disciplines, taking the operating system curriculum of the computer specialty as an example, the causes of the above will be deeply analyzed. Discuss the methods of teaching reform and practice about ideologies and politics from three aspects: distilling the ideological and political elements, integrating ideological and political contents and perfecting ideological and political assessments. In the implementation phase, the teaching group reforms traditional teaching methods, and designs rich teaching cases. Specific ideological and political elements binds into the knowledge points of the operation system curriculum. To maximize the effectiveness of ideological and political education, the link of assessments are further improved. In the semester of ideological and political teaching, the student participation and the quality of experimental report were significantly improved, and the excellent rate of performance was nearly 20%higher than that under the traditional teaching mode. The effect shows that by promoting the integration of computer specialty education and ideological and political education, the effect can be maximized.关键词：teaching reform;ideological and political education;operating system
- 41
- |
- 18
- |
- 0
发布时间：2022-03-28 - 摘要：The imbalanced classification is a hot and difficult topic in machine learning nowadays. In order to improve the classification effect of imbalanced datasets, propose an imbalanced classification algorithm which combines resampling methods and MetaCost—RS-MetaCost. First, resampling the imbalanced datasets before MetaCost subsets, that is, over sampling minority classes or under sampling majority classes, to reduce or eliminate the degree of data imbalance. Secondly, in the stage of prediction probability, m-estimation is used to increase the prediction probability of minority class, which increase the prediction probability of minority class. RS-MetaCost is compared with classical algorithms with 6 simulated datasets and 10 real-world datasets. The results show that RS-MetaCost can improve the classification accuracy of a few classes under the premise of ensuring the overall classification accuracy is very high on the most of imbalanced datasets. Furthermore, the over-sampled RS-MetaCost is better than the under-sampled RS-MetaCost.关键词：imbalanced classification;MetaCost;resampling;m-estimation
- 26
- |
- 13
- |
- 0
发布时间：2022-03-28 - 摘要：Video facial expression recognition is widely used in driverless technology,intelligent medical treatment and other fields. Aiming at the problem of information loss in single-frame feature extraction of video, a single-frame enhanced convolutional network is proposed, which uses the fusion of shallow features and deep features to achieve feature enhancement. The shallow features are the CNN epitaxial convolution module to achieve shallow feature extraction. The deep feature is the fusion of the dilated convolution and the inter-channel attention mechanism in the CNN network to realize the feature channel relocation and the combination of strong and weak information. Based on the correlation between adjacent frames of video, a multi-frame enhanced convolutional network is proposed, which introduces an attention mechanism to extract key frames. Finally, it was verified on the AFEW dataset, CK+dataset, SFEW dataset, and FER dataset. The accuracy rate on the AFEW dataset was increased from 40.00%to 45.19%, and the F1 score was increased from 0.31 to 0.3937.The network model can be applied not only to static images, but also to dynamic videos. At the same time, it can also improve the accuracy of facial expression recognition, reduce errors, and improve recognition efficiency.关键词：expression recognition;single-frame enhanced convolutional network;attention mechanism;multi-frame enhanced convolutional network;AFEW dataset
- 29
- |
- 13
- |
- 0
发布时间：2022-03-28 - 摘要：For existing body recognition models in terms of insufficient detection accuracy, long detection period and too large model parameters, propose a new improved posture recognition model—KP-Detector. The joint points and limbs are detected and identified separately.a new and improved PLF (Point Line Fields) and Dense connection mechanism is used in the model to reduce the complexity of the model and alleviate the disappearance of the gradient of the model. The Hungarian algorithm is used for efficient matching of the limbs, and the 6-layer model structure is optimized. Point positioning phase and two-layer limb detection phase), the model can be applied to single and multi-person joint point detection problems at the same time. On the MII data set, the test accuracy of this model is better than that of the comparison model, the test speed is nearly 4 FPS faster than other models, and the model size is only 18 M, which shows that this model has more advantages than other models.关键词：KP-Detector;body recognition;PLF;deep learning
- 46
- |
- 4
- |
- 0
发布时间：2022-03-28 - 摘要：Small target detection is a popular and difficult problem in the field of target detection in recent years. Aiming at the existing missed detection problems of small target detection and the problems with high hardware performance requirements, improves Tiny YOLOV3 and proposes a small target detection algorithm AE-Tiny YOLOV3 suitable for use on low-performance platforms. Firstly, this paper uses the EfficientNet-B0 backbone network to replace the feature extraction network of the original algorithm; secondly, add a detection branch to the detection network to make the original algorithm two-scale prediction form a three-scale prediction; finally, it introduces an attention mechanism to improve the three detection branches. ResultThe experimental results show that on the VOC07+12 dataset, the AE-Tiny YOLOV3 algorithm meets the requirements of real-time detection and has certain robustness, which can increase the mAP value by 16.89%at the highest. The AE-Tiny YOLOV3 algorithm is applied to the detection of insulator status in overhead transmission lines. The mAP reaches 86.53%, which is 15.27%higher than that of Tiny YOLOV3, which can meet the real-time detection of small target insulator status.关键词：small target detection;Tiny YOLOV3;attention mechanism;multi-scale detection;insulator status detection
- 24
- |
- 5
- |
- 0
发布时间：2022-03-28 - 摘要：In order to identify the reading of pointer-type meters in substations in a complex environment, a reading recognition method of pointer-type meters based on deep learning is proposed. First use the target detection algorithm YOLOv3 to detect the specific position of the meter and the meter scale value in the picture, and use the character recognition algorithm based on the LeNet-5 network to identify the specific value of the scale value; then use the semantic segmentation algorithm DeepLabv3+to segment the meter pointer area; finally use The angle method reader indicates the number. The experimental results show that the algorithm proposed in this paper can efficiently and accurately read the indications of the pointer in different lighting, weather and background environments, with an average reading error rate of less than 3.5%, which can meet the daily inspection tasks of substation inspection robots.关键词：pointer-type meter;object detection;semantic segmentation;character recognition
- 16
- |
- 9
- |
- 0
发布时间：2022-03-28 - 摘要：In order to meet the accuracy and real-time requirements of clinical epileptic seizure prediction, a seizure prediction method based on time domain and frequency domain feature extraction is proposed. The algorithm uses a data segment with a moving step of 1s and a window size of 5s for feature extraction, and replaces the original data for classifier training. The classification standard of the sample is that 15 minutes later, whether the patient has seizures or not, taking seizure as a negative sample. The classifier of the algorithm is LightGBM. The algorithm was applied to a test on the epilepsy data set of a hospital in Nanjing, and the test results are the recall rate of the training set on the epilepsy data set of a hospital in Nanjing was 100%, the false alarm rate was 0/h and the test set recall rate is 84.18%, and the false alarm rate is 0.57/h. The algorithm can solve the classification problem of existing data well, which has certain application value in the prediction of epilepsy.关键词：seizure prediction;EEG;LightGBM;decision tree;wavelet transform
- 14
- |
- 7
- |
- 0
发布时间：2022-03-28 - 摘要：Human activity recognition technology based on wearable sensors has been widely used in many fields. Extracting rich features from human activity signals is one of the keys to improve the accuracy of activity recognition. In this paper, a fusion feature based on Fourier descriptor (FDs), local binary pattern (LBP) and wavelet energy spectrum (WES) is proposed to extract detailed information of human activities. In order to improve the reliability of the recognition system and remove redundant features that have no impact on the recognition accuracy, Relief, a filtering selection algorithm, is introduced in this paper to select features, and features with high discrimination for different activities are selected. Finally, random forest (RF) classifier is used to accurately identify a variety of different activities. Based on Python3.6 platform, the effectiveness of the algorithm is verified on the public WISDM and ADL dataset. Experimental results show that the multi-feature fusion algorithm recognition accuracy of WISDM and ADL is 94.5%and 95.3%, respectively, which has better recognition performance than the single feature algorithm, and has strong robustness.关键词：Fourier descriptor;local binary pattern;feature selection;human activity recognition
- 9
- |
- 4
- |
- 0
发布时间：2022-03-28 - 摘要：In order to solve the problem that the video frame image is blurred due to road bumps and the vehicle's own jitter, which affects the vehicle detection effect. The optical flow method based on feature matching combined with the SURF feature point extraction algorithm is used to preprocess the vehicle video for anti-shake, and then the stabilized video sequence is transferred to the trained YOLOv4 frame for vehicle detection. The algorithm is verified on KITTI dataset, the final recognition accuracy is 96.5%.The experimental results show that the anti-shake optimization algorithm has obvious effects, and the ability to detect blurred frames in video sequences has been greatly improved.关键词：vehicle detection;video anti-shaking;target recognition;optical flow method
- 14
- |
- 3
- |
- 0
发布时间：2022-03-28 - 摘要：Adaptive weighted pooling was proposed aiming at the problem of spatial information loss and inaccurate extraction of important freatures in the pooling process in convolutional neural networks. In this method, a set of trainable weight parameters are set according to the size of the pooling window, and the weight parameters are weighted and summed with the sorted values in each pooling window to obtain the pooling result. In the process of network back propagation, the weight parameters are iteratively updated through gradient descent to obtain the optimal weight parameters. Different pooling methods are used for comparative experiments of image classification on the Fashion-minist, Cifar10 (using shallow CNN and ResNet18 network structures respectively), and Omniglot datasets. Adaptive weighted pooling increases the classification accuracy of the test set by 0.21%, 0.43%, 0.80%and 0.66%.The proposed adaptive weighted pooling enables the neural network to select different pooling strategies according to different task types, and achieves higher accuracy in image classification compared with common pooling methods.关键词：convolutional neural network;adaptive weighted pooling;image classification;downsampling
- 18
- |
- 12
- |
- 0
发布时间：2022-03-28 - 摘要：In order to solve the problem of estimating the safe social distance of pedestrians in surveillance videos with uncalibrated cameras, a method combining pedestrian detection, homography and scale estimation was proposed to classify whether pedestrians were in the safe social distance in monocular cameras. Firstly, based on the YOLOv5s network, the pedestrian detector with better robustness was trained by using the data of MSCOCO containing only pedestrians. Secondly, according to the assumptions of the camera imaging model, the homography matrix from the scene ground to the image plane was deduced, and the scale information from the scene ground to the local area of the image was estimated by the average height of human and the height of the image pedestrian detection box, so as to project the elliptical pedestrian safety area on the image. Finally, the overlap of the pedestrian safety area in the image is calculated to determine whether the safe social distance is violated. Experimental results show that when IoU=0.5, pedestrian detection precision, recall rate and AP of MSCOCO dataset reach 81.39%, 82.39%and 76.52%, respectively. The precision, recall rate and F1-score of pedestrian safety social distance classification of OTC dataset reached 98.99%, 89.12%and 93.79%, respectively. The proposed method has high performance in pedestrian detection and safe social distance violation detection under normal security camera circumstances.关键词：pedestrians social distance estimation;object detection;homography matrix;monocular vision
- 7
- |
- 3
- |
- 0
发布时间：2022-03-28 - 摘要：The development of strategic marine and marine economy has become an important strategy for national development with the rapid development of economy. The construction and improvement of harbor facilities play an important role in the development of marine economy. However, the operation of heavy machinery and marine machinery in the harbor scenes poses a potentialsafety risks to the operators, hence, the construction of intelligent and information-based harbors is imperative. In recent years, the rapid development of computer vision and deep learning technology has provided strong technical support for the application of intelligent vision technology in harbor scenes. In this paper, we build an operator detection platform based on the deep learning framework YOLOV4.We collect and sort out a large-scale harbor operator data set in a self-built harbor, and realize accurate detection of harbor operators in different operating scenarios on this data set. In our experiments, we compare three state-of-the art object detection frameworks, i.e., Faster RCNN, SSD and YOLOV4.Extensive experiments on the self-built harbor operator data set shows that the mean Accuracy Precision of YOLOV4 is better than other object detection frameworks. The YOLOV4-based harbor operator detection technology has improved the progress of informatization construction to a certain extent and improved the safety of harbor operators.关键词：harbor operation;object detection;YOLOv4;computer vision
- 35
- |
- 12
- |
- 0
发布时间：2022-03-28 - 摘要：Temperature is one of the most important parameters affecting climate, and temperature forecasting is of great significance for identifying extreme meteorological disasters such as droughts and floods. Based on the machine learning theory, a multi-information fusion temperature forecasting method combining random forest (RF) and one-dimensional convolutional neural network (1DCNN) is proposed. Firstly, use difference method to transform meteorological observation data into stationary time series data; Secondly, use the RF model to dig out feature variables that are highly related to temperature as the input variables of the neural network model; Finally, build a multi-information fusion temperature forecast model RF-1DCNN. Taking the historical meteorological observation data in Kunming, Yunnan Province as an example, compared with the traditional LSTM, 1DCNN and Back Propagation Neural Network (BP) of the temperature forecast in the next 10 hours is carried out. The research results show that compared with LSTM, 1DCNN and BP, the root mean square error (RMSE) of 1DCNN is reduced by 13.110%, 26.176%and 17.612%, and the Pearson correlation coefficient (r) is reduced by 0.240%, 0.567%and 0.355%, respectively. This research method has good learning ability and generalization ability, and provides a reference basis for accurate temperature forecasting.关键词：RF;1DCNN;multi-information fusion;temperature prediction
- 17
- |
- 7
- |
- 0
发布时间：2022-03-28 - 摘要：Non-intrusive load identification is the first step of advanced power measurement system, which is of great significance to the construction of smart grid. In order to solve the problems of slow speed and low accuracy of traditional non-invasive load identification algorithm, a non-invasive load identification method based on sparrow search algorithm (SSA) to optimize extreme learning machine (ELM) is proposed. The SSA is used to obtain the optimal input weights and thresholds of the hidden layer of extreme learning machine, and the SSA-ELM non-invasive load identification model is constructed. The model is put on six commonly used family load data sets to carry out load identification experiments. The results show that the recognition accuracy of non-invasive load identification algorithm based on SSA-ELM is 96.1%, which is better than that of traditional ELM and BP neural network algorithms (86.3%and 91.8%) .The non-invasive load identification algorithm based on SSA-ELM can be effectively applied to the non-invasive load identification of household electricity load.关键词：non-intrusive load identification;extreme learning machine;sparrow search algorithm;parameter optimization
- 11
- |
- 5
- |
- 0
发布时间：2022-03-28 - 摘要：In order to effectively collect point cloud data inside karst caves and overcome the problem that a large number of data is missed in obtaining point cloud data inside karst caves by a single data collection method, introduce the result by using the ground-mounted 3D laser scanner, backpack 3D laser scanner and photogrammetry technology to obtain multi-source point cloud data in karst caves, which are fused by the improved ICP algorithm then. The results showed that the use of multiple data collection methods can effectively collect three-dimensional point cloud data in karst caves, compensate for the lack of data caused by a single data collection method, and obtain relatively complete point cloud data. By using the improved ICP algorithm, better data fusion effects can be obtained.关键词：karst cave;three-dimensional laser scanning;photogrammetry;multi-source point cloud data;point cloud fusion
- 10
- |
- 4
- |
- 0
发布时间：2022-03-28 - 摘要：In order to solve the problem that the sending request of high priority message always blocks the sending request of low priority message in CAN bus, that is, the strategy based on fixed priority in CAN bus may lead to the starvation problem of low priority message, a CAN bus high priority inversion algorithm to avoid starvation is proposed. The algorithm uses a non-preemptive rate monotone algorithm to allocate the priority of CAN bus messages. The average response time and arrival rate of low priority group messages under different can bus utilization are analyzed by simulation. The results of the running simulation showed that after using the CAN bus high priority inversion algorithm to avoid starvation, as the bus utilization increases, the average response time of the message is reduced by 6.9%, 9.5%, 10.7%, 21.9%and 43.29%respectively, and when the bus utilization exceeds 100%, the arrival rate of low-priority group messages can still be 1.Therefore, the proposed algorithm can effectively avoid the starvation problem caused by low priority messages on CAN bus, which proves the effectiveness of the algorithm.关键词：priority inversion;CAN bus;priority queue;response time
- 12
- |
- 4
- |
- 0
发布时间：2022-03-28 - 摘要：In order to better dig out the inherent correlation characteristics of tobacco redrying formula to maintain the redrying formula more effectively, Apriori algorithm needs to manually preset the minimum support and confidence, and there are a lot of redundant rules, an improved binary particle swarm optimization (LDABPSO) is proposed to mine the associated features of tobacco redrying formula. Aiming at the problem that binary particle swarm optimization is easy to mature prematurely and fall into local optimum, an improved strategy is proposed from three aspects: population initialization, stagnation of feasible solutions, and disturbance mechanism; Secondly, compare the performance of LDABPSO algorithm with BPSO algorithm and ABPSO algorithm on 6 standard test functions; Finally, the formula data of Yunnan Qilin Redrying Plant in recent years is used as the data source for association rule mining, and the LDABPSO algorithm is used for association rule mining. The experimental results show that when the minimum support and minimum confidence of the Apriori algorithm are set to 0.18 and 0.5 respectively, the number of tobacco rules mined by the LDABPSO algorithm is reduced by 90.23%and the running time is reduced by 8.4%compared to the Apriori algorithm, proving the effectiveness of the algorithm.关键词：redrying formula maintenance;association analysis;LDABPSO algorithm
- 12
- |
- 5
- |
- 0
发布时间：2022-03-28 - 摘要：In order to overcome the difficulty of inefficient use of the limited spatial space at a signalized intersection, an unsignalized traffic management model is introduced based on timed automaton. Firstly, the intersection is divided into several disjoint spatial traffic resources, and the dynamic process for each vehicle passing through the intersection is modeled by a timed automaton. Then, the overall system model is generated by paralleling all the constructed timed automata. A vehicle scheduling strategy for maximizing the throughput of the intersection is designed under the premise of preserving safety. The experimental results show that, compared with results obtained using the traditional signal control method, the unsignalized control method introduced makes better use of the spatial space of the intersection, and the vehicle throughput capacity at the intersection is further improved.关键词：timed automaton;unsignalized intersection;resources scheduling;traffic control
- 9
- |
- 4
- |
- 0
发布时间：2022-03-28 - 摘要：Miniaturization is the development trend of Raman spectrometer. Area array CCD, which has the advantages of high sensitivity, large dynamic range, high quantum efficiency and small volume, has become a research hotspot in the field of Raman spectrometer. In this paper, an area array CCD spectrum acquisition system based on field programmable logic device is designed. The system uses S11510-1106 area array CCD of Hamamatsu company as photoelectric sensor, controls the whole system with Verilog language through FPGA chip, and uses CCD special signal processing chip AD9823 and analog-to-digital converter ADS8381, the CCD signal is denoised and the high-speed A/D conversion is realized based on the principle of correlation double sampling. At the same time, the CCD data is transmitted through the serial port. Finally, the collected spectral data is displayed on the host computer software. The test results showed that the design can correctly collect and display spectral information, and has the advantages of small circuit volume, accurate control timing and good acquisition effect, which can be used as the spectral acquisition system of micro Raman spectrometer.关键词：Raman spectrometer;spectrum acquisition;area array CCD;FPGA;correlation double sampling;Verilog
- 18
- |
- 7
- |
- 0
发布时间：2022-03-28 - 摘要：In order to reduce the operation difficulty of the underwater remote control robot and improve the controllability and operability of the robot, the operating system of the underwater robot is developed by using Java. Firstly, the hardware structure of underwater robot monitoring system is introduced, and the functional structure of underwater robot monitoring software developed by Java is put forward. The composition and functions of monitoring module, control module, feature information module and alarm module are introduced. The communication mode and protocol of monitoring system software are described, and the software flow of monitoring system and main control board are introduced respectively. Experiments show that the underwater robot monitoring system based on Java can coordinate all modules well, and has the characteristics of short development and good expansibility.关键词：underwater robot;monitoring system;Java;system framework;sensing and control
- 28
- |
- 7
- |
- 0
发布时间：2022-03-28 - 摘要：With the increasing demand of LCD, it is very important to improve the production efficiency of LCD. In the process of LCD automatic assembly, the clamping and positioning device is an important equipment in the automatic production line. In this paper, aiming at the display model of 215a, 215p, 238A and 240A, a kind of centering positioning fixture is designed. According to the principle mechanism of the designed centering positioning fixture, the mathematical modeling is carried out, and the specific parameters of each component of the centering positioning mechanism are completed. The constraint equation is solved by MATLAB, and the crank length of the symmetrical crank connecting rod mechanism in the centering positioning fixture is 85.604 7mm, and the connecting rod length is 279.986 0mm, which provides a certain theoretical basis for the actual production assembly.关键词：liquid crystal display;alignment positioning;fixture;mathematical modeling;MATLAB
- 14
- |
- 5
- |
- 0
发布时间：2022-03-28 -
Design of Circuit System of Ultrashort Wave Physiotherapy Instrument Based on Microcontroller Unit AI导读
摘要：In order to realize the digitization and automation control of ultrashort wave fields, this paper designs a digital control can be automatic tuning of ultrashort wave fields circuit system, the system by pushing from type of self-excited oscillation circuit ultrashort wave frequency is 40.68MHz, based on microcontroller unit as main control chip step-type tuning method control variable capacitance matrix power output circuit automatic tuning. The system mainly includes the power module, the ultrashort wave generation module, the microcontroller unit control module, the input display module, the automatic tuning module, the ultrashort wave output module, and the digital control is realized by the microcontroller unit. After the whole machine is powered on for testing, it can realize the digital display of treatment intensity and treatment time, produce a stable frequency of 40.68MHz ultrashort wave, can produce 22~48℃thermal effect, and realize the digitization and automation of the instrument.关键词：ultrashort wave therapy instrument;digital control;auto tuning;microcontroller unit- 15
- |
- 7
- |
- 0
发布时间：2022-03-28 - 摘要：With the continuous development of society and changes in people's lifestyles, cardiovascular diseases has become an important cause of death. In order to make effective and rational use of medical resources. Artificial intelligence methods is used to construct the knowledge map of cardiovascular diseases, and an automatic question-and-answer system of cardiovascular disease is developed based on the map. Experiments show that the precision rate and recall rate of the system answer are 0.95 and 0.93, and the F1 value is 0.94, which can effectively answer the relevant questions of users in the diagnosis of cardiovascular disease symptoms and drug recommendation, and save medical resources to a certain extent.关键词：cardiovascular diseases;knowledge graph;question answering system;artificial intelligence
- 23
- |
- 13
- |
- 0
发布时间：2022-03-28 - 摘要：With the development of Industrial Internet in recent years, industrial micro-service has become a very popular topic. Traditional modeling mostly uses Simulink and other tools, and the reuse of the established model is poor, and it cannot be run without the simulation platform. In view of the above situation, use OpenModelic to conduct PMSM modeling, uses Python and Java languages to make encapsulation and call to the FMU file exported by the built model, and forms the motor model micro-service, which is finally deployed on the Spring Cloud architecture. The experimental simulation of the motor model micro-service shows that it is feasible to transform the traditional model into micro-service, and the design improves the reusability of the model, which can make the model run off the platform.关键词：OpenModelica;Spring Cloud;PMSM;V/F controller;DC/AC inverter;microservice
- 12
- |
- 3
- |
- 0
发布时间：2022-03-28 - 摘要：Aiming at the problem that the current network abnormal traffic detection methods pay little attention to the time dependence of network traffic and do not detect network abnormal traffic from the perspective of time cycle, this paper proposes an unsupervised network abnormal traffic detection method based on CGAN-LSTM. This method uses the generator and discriminator of LSTM structure to learn the data characteristics of normal samples, uses time period information to guide generator to generate samples, and finally uses the reconstruction error of generator and the discrimination result of discriminator to discriminate test samples at the same time. The experimental results show that the F1scores of this algorithm on ISCX2012 data set and CICIDS2017 data set reach 89.38%and 85.62%respectively, which has better detection performance compared with the existing unsupervised abnormal traffic detection algorithms.关键词：anomaly detection;unsupervised;periodicity;CGAN;LSTM
- 14
- |
- 3
- |
- 0
发布时间：2022-03-28 -
Comparison and Improvement of Darknet Traffic Detection Based on Deep Convolutional Neural Network AI导读
摘要：In order to study the applicability of large deep convolutional neural networks in the field of multi-class darknet traffic detection, the classification performance were compared based on Resnet, Densenet and Xception networks, respectively. Put the three models on the Darknet2020 data set for verification, using 93 000 pieces of Non-Tor data, 1 300 pieces of Tor data, 23 000 pieces of Non-VPN data and 22 000 pieces of VPN data for experiments. The results show that the three models can quickly process massive data, and the detection results of Tor and Non-Tor traffic are good, with the F1 value up to 0.91, while the detection results of VPN and Non-VPN need to be improved. Among them, the Densenet network has the best detection performance on the test set. After adding the GRU network to extract the timing features for improvement, the overall classification accuracy rate is 83.4%, the recall rate is 82.2%, and the detection performance is further improved.关键词：convolutional neural network;darknet traffic;Resnet;DenseNet;Xception;GRU- 12
- |
- 4
- |
- 0
发布时间：2022-03-28 - 摘要：In order to improve the problems of traditional fuzzy C-means (FCM) clustering algorithm in SAR image segmentation, such as the large number of iterations, low robustness, and poor segmentation accuracy, propose an improved FCM algorithm by selecting the main pixels and combining the non-local information of the image. The algorithm first divides the image into blocks and then selects main pixels to form a main pixel set, and then uses K-means clustering on the pixel set to determine the initial clustering center, and then introduces non-local spatial information into the objective function of the FCM algorithm, and uses grayscale The information and spatial information cooperate with the adaptive smoothing factor to enhance the anti-noise performance of the algorithm, and finally use the image morphology operation to complete the image segmentation. Experimental results show that the algorithm proposed in this paper has high segmentation accuracy for simulated and actual SAR images, reaching 99%and 96%respectively. Compared with the traditional algorithm, it not only improves the segmentation accuracy, reduces the number of iterations, but also has good anti-noise Performance, while retaining the details of the image.关键词：FCM clustering;SAR image;non-local information;image segmentation
- 13
- |
- 5
- |
- 0
发布时间：2022-03-28 - 摘要：An active contour model that combines local information was proposed to improve the accuracy and efficiency of traditional noisy images segmentation. Firstly, a new velocity function based on local image information was constructed, which makes the velocity function change with the local gray information of the image; Secondly, replace the original image with the fitting image and increase the weight of the fitting center to construct a new fitting term; Finally, adaptive weight coefficients are used to increase the flexibility of the model for segmentation of different images. Compared with C-V model, the segmentation time of three noisy gray images is improved by 72.30%, 82.95%and 75.79%respectively. The experimental results show that when the traditional C-V model, LCV model and RSF model are used to segment the noisy gray image, there are errors in segmentation. The proposed model can accurately segment noisy grayscale images. These segmentation results are not over-segmented for the background and foreground, and the segmented contour edges are smooth.关键词：noisy image segmentation;active contour model;velocity function;adaptive weight coefficients
- 12
- |
- 5
- |
- 0
发布时间：2022-03-28 - 摘要：To address the problems of low grayscale in some regions of PET and MRI medical images after fusion, which are prone to artifacts and insufficient feature extraction with missing details, we first design an image fusion method based on NSST multiscale transform, design a residual network for high frequency subbands to fuse PET and MRI images after weight assignment, and fuse low frequency subbands according to the fusion rule of sparse representation. Finally, the fused results are obtained by inverse conversion. After that, a dense convolutional network framework is designed to introduce a self-attentive mechanism to assign weights to features in different regions of the image, introduce scale-varying layers between dense blocks to obtain feature maps at different scales, and fuse feature maps at different scales across layers to recover the information lost due to multi-layer convolution. To recover the feature details lost due to pooling, the post-processor module postProcessor is designed to process the feature maps extracted from the multi-scale layers. Finally the model obtains the classification labels through a fully connected layer with 2 neurons with a softmax classifier. The model was experimented on the Harvard Whole Brain Atlas dataset with TCIA, and the results showed that the fused PET-MRI images of horizontal brain sections could better reflect the central sulcus sign as well as the limbic branch sphincter sign with 97.6%and 98.2%classification accuracy.关键词：PET/MRI;glioma;medical image fusion;DenseNet;self-attention mechanism;NSST
- 14
- |
- 5
- |
- 0
发布时间：2022-03-28 - 摘要：To solve the problems of low precision, weak anti-interference ability and poor robustness in existing binocular distance measurement method, a binocular distance measurement method based on image feature matching using deep learning is proposed. Firstly, images from the binocular camera go through a self-supervised keypoint detection network. The keypoints and descriptors can be obtained by two decoders. Then, the keypoints are matched with the descriptors and compute disparity. Finally, true distance is obtained according to the principle of similar triangles of binocular camera. Experimental results on KITTI datasets show that the proposed feature detection network demonstrates better performance than traditional methods and other methods based on deep learning. Experimental results on binocular distance measurement show that the average error of the proposed method is 0.58%within the range of 20~150cm, which means the proposed method can meet the requirements of accurate ranging.关键词：binocular distance measurement;deep learning;feature matching;self-supervised training
- 23
- |
- 9
- |
- 0
发布时间：2022-03-28 - 摘要：With the renewal and iteration of electronic products, the requirements for colloid quality of electronic product parts are constantly improving. For reinforcement glue recognition and segmentation, the traditional algorithm has low robustness, and the deep learning semantic segmentation network Unet segmentation speed is slow. Therefore, a real-time semantic segmentation network fast Unet based on improved Unet is proposed. The network has three feature extraction branches. The output feature graphs are the size of the original graph, 1/4 of the original graph and 1/16 of the original graph respectively. Each branch shares a part of the network weight. In the second feature extraction network, channel segmentation, convolutional block attention module and pyramid pooling module are added. The experimental results show that compared with Unet network, the proposed fast Unet network is improved by 0.07 in both MIoU and MPA, 43.08 in FPS, and the online detection time of a single sample is only 25ms, which significantly improves the segmentation effect of reinforcement colloid online detection.关键词：real-time semantic segmentation;Unet;reinforcing glue;attention module;pyramid pooling module
- 10
- |
- 4
- |
- 0
发布时间：2022-03-28 - 摘要：Discuss the use of deep learning methods to solve the uncalibrated photometric stereo technology of non-Lambertian surfaces. Existing photometric stereo algorithms usually assume that the light source conditions are known or simplify the reflectance model, but this limits its practical application. Therefore, propose an uncalibrated photometric stereo technology based on multi-scale aggregation to generate a confrontation network, which directly learns the mapping relationship between photometric images and normal maps, avoiding cumbersome light source calibration steps and accurately estimating the expression of reflectance function. Experimental results show that the method proposed is more robust than the existing optimal uncalibrated photometric stereo vision algorithm, and the average angle error is reduced by 1.92°, MSE is reduced by 1.43%.Several ablation experiments have proved the superiority and effectiveness of the network structure.关键词：uncalibrated photometric stereo technology;multi-scale;generative adversarial network
- 11
- |
- 3
- |
- 0
发布时间：2022-03-28 - 摘要：Knowledge graph question answering systems (KGQA) has been widely used in many fields such as medical care, finance and e-commerce, because of its precise and efficient ability. Recently, with the rapid development of knowledge graph complete (KGC) and deep learning (DL) technology, KGQA methods have been forward continuously. In accordance with the development of KGQA method, the KGQA research is summarized. First, introduce three KGQA methods including semantic parsing-based, information retrieval-based and knowledge embedding-based methods. Second, the general domain datasets and specific domain datasets in the evaluation task of KGQA commonly used be introduced in detail. Finally, the challenges of KGQA methods and future research directions are summarized.关键词：knowledge graph;question answering system;semantic parsing;information retrieval;knowledge embedding;neural network
- 26
- |
- 7
- |
- 0
发布时间：2022-03-28 - 摘要：Camouflage object detection (COD) is a new research field, its goal is to detect those objects hidden in the scene image background, which can be applied to species protection, military monitoring and etc, has a wide range of application area. Firstly, the basic concept of camouflage detection is introduced. According to the key technologies adopted, COD methods can be divided into two categories: traditional camouflage detection methods based on hand-crafted features, and deep learning based camouflage detection methods. Both the categories are analyzed and discussed in detail, including the principle and technical details. Secondly, the corresponding data sets and performance metrics are also introduced, and the representative COD algorithms are compared and analyzed qualitatively and quantitatively. Finally, pointed out the problems remained to be resolved and the future research directions of camouflage detection and segmentation.关键词：camouflage detection;deep learning;attention mechanism;convolution network
- 17
- |
- 5
- |
- 0
发布时间：2022-03-28 - 摘要：With the increase of flight load and flight time, the unsafe events caused by flight fatigue happen occasionly. The research on the quantification, evaluation, detection, prevention and recovery of human fatigue is of practical significance. The fatigue detection method based on neural network has been used more early in landtransportationfields, but less in civil aviation. Therefore, the migration of fatigue detection methods based on neural network to civil aviation is of great significance to promote the development of fatigue detection methods in civil aviation. The traditional flight fatigue detection method and its characteristics, the application of neural network in flight fatigue detection are described. Finally, the development trend of flight fatigue detection method is prospected.关键词：flight fatigue;neural network;fatigue detection;aviation safety
- 13
- |
- 6
- |
- 0
发布时间：2022-03-28 - 摘要：Pancreatic cancer is an insidious malignant tumor of the digestive tract with a very poor prognosis. Pancreatic medical imaging is an important tool for doctors to diagnose pancreatic diseases. Deep learning has been widely used in image, language and other fields, and its application in the field of medical imaging has also become a research hotspot. The history of the development of deep learning was reviewed, the application of deep learning in pancreatic pathology images, CT images, and ultrasound images was explained, and the prospects for future development were made.关键词：deep learning;medical imaging;pancreatic cancer;pathology image;CT image;ultrasound image
- 16
- |
- 3
- |
- 0
发布时间：2022-03-28
- 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