Abstract:Collaborative development technology is an important method for the development of control system software, and the integrated deployment of software is an important part of collaborative development technology. With the increasing scale and complexity of software, the existing methods of software integration deployment have the problems of low efficiency and low security due to the lack of standardized and unified deployment specifications and security mechanisms. Aiming at the above problems, a collaborative integration deployment method based on metric authentication is proposed to solve the security and efficiency problems brought by integrated deployment work in distributed collaborative development environment. The main idea of this study is to use a security association protocol to establish a trusted secure transmission channel between the software deployment parties, and then complete the software integration and deployment according to the unified package deployment specification. Theoretical analysis and experimental results show that the scheme unifies the software package deployment specification, optimizes the integrity measurement algorithm, and reduces the time for software authentication and deployment. Since the two parties establish security associations, the security of the software deployment process can be improved. The research content of this study can complete safe and efficient software integration deployment in a distributed development environment.

Abstract:As an emerging technology, Blockchain has gained much attention due to its decentralization, transparency, and non-tamperability. Because the study to Blockchain is still in its early stages, there are still a number of data management issues in the application development process, such as data privacy, scalability, and latency. In this study, we propose the design patterns of Blockchain data management, namely, data privacy protection pattern, hash integrity pattern, and state channel pattern, to help developers use Blockchain for application development. All the above patterns are implemented and verified in a Blockchain-based traceability system: originChain, to show the effectiveness of the proposed patterns.

Abstract:The Flight Deck Aviation Support Job (FDASJ) on the large ship cover all the activities happening during the period from the airplane landing on the ship to the airplane taking off from the ship. It has been a challenge problem to detect the FDASJ based on videos. The difficulty is to detect multi-FDASJs occurring in the same scene during the same time period, participants of which are located alternatively. In this study, we propose an FDASJ detection method based on video event. The FDASJ is represented as the complex event composed of the fine-grained FDASJs. The FDASJ structure extraction algorithm based on trajectory hierarchical clustering is first adopted to determine the hierarchical combination relationship between FDASJs and the participators of each FDASJ; the FDASJ identification algorithm based on activity model is performed to identify the FDASJs. Experimental results show that the proposed method achieves sound performance and could detect multi-FDASJs occurring in the same scene during the same time period, participants of which are located alternatively.

Abstract:This study models the abnormal workflow in hospitals by extending the components of polychromatic sets and polychromatic graph. First, the research progress of abnormal workflows is introduced. Second, the theories needed for formalization of this study—polychromatic sets and polychromatic graph are described. According to the particularity of the research object, the components in the polychromatic sets are expanded to construct personal colors and the enclosure matrix of the volume represents the relationship between anomalous events and anomaly types. The enclosure matrix of edges and objects has been added to the polychromatic diagram to describe the starting point and ending point of the arc of abnormal handling behavior, constructed an hospital abnormal workflow processing model based on polychromatic graph. Finally, the hospital outpatient registration process exception handling is used as an example to verify the correctness of the model.

Abstract:In view of the slow loading speed of cable 3D visualization scene, a simplified algorithm for extracting the outer surface of 3D model and a multi-level of detail R-tree index data scheduling organization method are proposed. Firstly, the LOD level simplification of cable well and pipe trench models occupying a large amount of memory in three-dimensional scene is carried out, and the experimental results show that the amount of data is greatly reduced. Then, the simplified data is organized and scheduled according to the R-tree index structure of multi-level of detail. Compared with the traditional R-tree, the R-tree constructed by this method constructs a better tree shape in node selection and node splitting, which makes the data in progress. In the process of indexing and scheduling, it can improve the loading speed of 3D scene of cable engineering, and effectively realize the smooth display of 3D model in cable engineering.

Abstract:Named Entity Recognition is a key technology in natural language processing, and the methods based on deep learning have been widely used in Chinese entity recognition. Most deep learning models focus on the feature extraction of words and characters, but ignore the semantic information of word context, therefore, they cannot represent polysemy, and the performance of entity recognition needs to be further improved. In order to solve this problem, this study proposes a method based on the BERT-BiLSTM-CRF model. First, word vectors based on context information are generated by the pretreatment of BERT model, and then the trained word vector is input into BiLSTM-CRF model for further training. The experimental result shows that the proposed model achieves sound results and reaches F1-score of 94.65% and 95.67% respectively in the MSRA corpus and People’s Daily.

Abstract:It is of great significance for researchers to master the development of research field through the analysis of published papers. In order to meet this requirement, a visual analysis method based on extended B_cp index was proposed. First of all, keywords containing phrases are automatically extracted from the title, abstract, and author provided keywords. Then co-occurrence relationship between these keywords was extracted. According to these keywords, LDA algorithm was used to extract topics. Then, an extended B_cp index was proposed to measure the development state of keywords. Based on this method, a visual analytic tool VISExplorer was designed and implemented. VISExplorer can show the distribution and development trend of domain topics, recommend high-quality papers, and browse top authors. Finally, taking the domain of visualization as an example, VISExplorer was conducted in real cases of publications on IEEE VIS Conference from 1990 to 2018, and the usefulness and effectiveness are proved by user’s feedbacks.

Abstract:Rule-based CPS monitoring methods have significant advantages in reducing monitoring complexity and improving monitoring flexibility. Currently, rule-based CPS monitoring methods do not consider the timing constraints of CPS monitoring scenarios, and only use various optimization techniques to shorten the response time. Based on the real-time rule engine, a real-time monitoring system of CPS, RTCPMS, is established. The system uses Rete network to represent the monitoring rules, and its core is a new real-time inference algorithm Rete-TC. The Rete-TC algorithm introduces the rule deadline, and the timing constraints of CPS monitoring is satisfied as much as possible by the priority-based Beta node scheduling method. The simulation experiment and smart building application case verify the effectiveness of the RTCPMS, and the experimental results show that the core algorithm Rete-TC has a better scheduling success ratio than the traditional rule inference algorithm Rete.

Abstract:With the development of scientific research, there is a massive increase in scientific research data. PB-level scientific research data requires efficient and stable storage systems. The traditional data storage scheme has problems such as poor resource utilization, low cluster expansion performance, and unfriendly user interface operation, which seriously limit the effective use of data. Relying on the Big Data Project of the Chinese Academy of Sciences, we design and implement an efficient storage system i-Harbor. Its core architecture is based on object storage system, using open-sourced Ceph distributed system and MongoDB database as the storage carrier of object data and metadata. The data interface is designed on the basis of HTTP API and FTP. To improve the platform disaster tolerance and security, we use Multiple Copies and Erasure Coding technology to eliminate single node of failure. Meanwhile we locate the real-time platform parameters and faults by Zabbix cluster monitoring system. Based on the distribution characteristics, the cluster can add storage nodes at will, which improves the platform’s scalability.

Abstract:In the process of disassembly and assembly of mechanical parts, there are problems such as large equipment investment, high loss, high risk, and low disassembly efficiency, this study uses mixed reality technology to design a virtual disassembly and assembly system. Among them, the disassembly logic is the core of the disassembly system, and the universality of the disassembly system depends on the design of the disassembly logic. In order to solve the problem of poor universality of the virtual disassembly system, this study analyzes and summarizes the connection and assembly relationship of common mechanical models, and designs a graph-based traversal algorithm. The data structure is used to express the logical relationship of each model, and experiments are carried out on the disassembly and assembly of special valves and six-joint manipulator. The experimental results show that the disassembly and assembly logic solved the problem of definability and versatility of the disassembly and assembly sequence, and can perform virtual disassembly and assembly of the model components according to the process flow, and finally realize the versatility of the virtual disassembly and assembly system.

Abstract:RapidIO protocol, as one of the data communication protocols, plays an important role in the development of embedded systems. It is suitable for short-distance application scenarios that require cooperation of multiple processing units, such as the board card system composed of multiple DSPs. As a high-performance digital signal processor, BWDSP chip has great potential in radar signal processing. In the hardware design and development, it is difficult to adapt the specific hardware resources directly using the existing data communication protocol, resulting in the low data transmission performance of the final product. Therefore, it is necessary to combine the specific hardware model to design the simulation model of data communication exchange model, so as to improve the efficiency of data transmission. This study first introduces RapidIO protocol and BWDSP architecture, then designs a serial RapidIO exchange model based on SystemC language, and finally designs and implements BWDSP virtual platform. The functions of BWDSP virtual platform designed in this study are in line with the actual RapidIO protocol standard, which has a certain guiding significance for the development of hardware products.

Abstract:Efficient transmission of message data is a research focus of hybrid networks. Publish/subscribe is the messaging pattern to realize the decoupling between publisher of messages and subscriber of messages, which is suitable for message data transmission between hybrid networks. By applying the publish/subscribe model to message data exchange, the format of message data is standardized, flexible management of various communication devices, and dynamic data routing based on message content. Along with publish/subscribe, a dynamic load balancing algorithm based on the round-robin scheduling is proposed to improve performance of low-rate networks. Simulation experiments are carried out, and the experimental results show that publish/subscribe model achieves reliable message data transmission between hybrid networks, and it performs better with load balancing.

Abstract:With the strong support of China for technology transfer, more and more technology transfer platforms have been emerging. However, it is difficult to match supply and demand on existing platforms. Therefore, the platform’s potential in improving the success rate of technology transfer is limited. For solving this issue, this study proposes a novel implementation method for technology transfer platform which uses big data mining technology to organize and analyze platform data. We combine other technologies such as full-text search, data collection, and RESTful interface to jointly improve the matching degree of the data, expand the visible range of the data, and ultimately improve the technology transfer rate. At present, a platform has been deployed and operated in Jiangsu Province. Through the platform, a large number of technology transfers have been implemented.

Abstract:The traditional network architecture often causes network interruption during the session process, which cannot provide sufficient security. The existing system cannot acquire and assemble the call record, and the telephone terminal cannot timely and reliably query the employee’s call record. An IMS telephone call sorting method based on the reliable routing protocol of the anchor node is proposed. The traffic capture module performs port mirroring on the captured traffic and parses the SIP protocol data. The current IP packet of the traffic passes the predicted link lifetime. The existence of candidate relay nodes to identify more reliable links for forwarding, to reduce the frequency of data loss, and improve the administrative efficiency of employees. The optimized routing algorithm RRCP proposed in this paper is compared with the traditional GRP algorithm. The network communication model is built by network simulation technology. The number of faults, packet loss rate, data delivery delay and average hop of the route optimization protocol are analyzed and verified. The number shows better performance.

Abstract:Computer-Assisted Interviewing (CAI) has been widely used in social survey research. Comparing with the general CAI system, this questionnaire survey system based on kinship network focuses on the design of the complex questionnaire, the construction of kinship network, and the analysis of the collected data. By using XML to express and store the complex questionnaire and answer data, this system provides researchers an efficient tool to design the complex questionnaire for in-depth interview and to localize the questionnaire into multiple languages so that it can be adapted to different cultural areas. An iterative algorithm is proposed for generating the diagram of kinship network. Based on this diagram, interviewers can efficiently and easily collect data with visualization interface of interviewing. The data analysis pre-processing generates the data matrices derived from the survey data in order that they can be easily imported into any statistical analysis software for further analysis. The proposed system can be employed in the research of family population structure analysis and social interaction between relatives and friends. Currently, this system has been utilized in several fieldwork sites in China and other countries and it satisfies the needs of social survey research.

Abstract:Aiming at the single information problem of traditional sentiment classification method, a multi-head attention model with user and product information is proposed. Firstly, hierarchical multi-head attention is used to replace single-head attention and obtain effective information from multiple perspectives. Secondly, using multi-head attention with user information and product information, mining the performance characteristics of user and product information in multiple subspaces, the model can get a more global impact of user preferences and product characteristics on sentiment score in multiple subspaces. Experimental results on IMDB, Yelp2013, and Yelp2014 datasets show that the performance of the proposed model is better than the other advanced baselines.

Abstract:The key to solve the problem of tracing the source of sudden water pollution in rivers is to determine its discharge time, location, and total amount. This study proposes two models which can quickly and accurately determine these three factors of heavy metal pollution sources and obtain the list of pollution sources through GIS. This study determines the dynamic analytical solution for the spatiotemporal changes of heavy metal pollutants in one-dimensional rivers, through the study of the hydrological and water quality of rivers and the characteristics of heavy metal pollutants. At the same time, a spatial-temporal traceability model of pollutants and a total pollutant discharge model are constructed, and this models are solved using the improved AFSA algorithm. The research results show that the algorithm enables the model to obtain the results of the three parameters more quickly and accurately, and then passes the results through the proposed method and uses GIS technology to provide the relevant staff with a probabilistic checklist of pollution source enterprises more quickly and accurately. The methods and models proposed in this study have certain guiding significance for water pollution treatment and protection of the water environment.

Abstract:In view of traditional feature selection methods such as information gain algorithm have preference for selecting features that have more values, Pearson correlation coefficient alone cannot be used to deal with nonlinear correlation, and optimization of algorithm parameters is too tedious, a feature selection fusion approach is proposed based on maximum information coefficient and Pearson correlation coefficient. Moreover, this approach makes use of genetic algorithm to optimize parameters automatically. In the first stage, the feature selection is carried out according to the maximum information coefficient and the correlation between features and tags. In the second stage, Pearson correlation coefficient is used to reduce the redundant acquired features. Furthermore, two hyper-parameters in the first two stages are optimized automatically based on genetic algorithm. The experimental results show that the algorithm can reduce the dimension of feature space and improve the classification performance.

Abstract:Light Guide Plate (LGP) marker detection is a crucial procedure in LGP manufacturing quality control but a mass of bubbling, polluted and marker-free cases may arise out of detection in traditional image algorithms. Because the mass of bubble, polluted and unmarked lines makes people difficult to design features. Compared with other classification neural networks, DenseNet Convolution Neural Network (CNN) has fewer parameters and stable gradient convergence. Because DenseNet CNN uses the idea of feature fusion, the accuracy of image classification is guaranteed. Through the transform learning method, the weights of the trained DenseNet network are transferred to the bilinear-CNN algorithm for training, which improves the local attention of the convolutional neural network and improves the accuracy of image classification. The implementation results show that the proposed method is feasible. Compared with the V2-ResNet-101, the accuracy of proposed approach is increased to 95.53%, while parameter number is decreased by 97.2%, and average single image detection time drops by 25% in the proposed network structure.

Abstract:In order to avoid the uneven energy consumption and energy hole in wireless sensor networks, this study proposes a data forwarding algorithm between clusters based on load division. Firstly, this study analyzes the problem that data forwarding between clusters may lead to energy hole near the sink. Then, the load allocation weight of cluster head is designed to represent the cluster head’s current energy state, data amount and distance. According to the load allocation weight, this study designs a new data forwarding algorithm, which forwards the data to cluster head near the sink. The simulation results show that compared with the basic EEUC algorithm and the classical LEACH algorithm, the proposed data forwarding algorithm has no great increase of energy consumption, prolongs the network survival time, and improves the efficiency of network energy utilization.

Abstract:The Bidirectional RNN (BRNN) was adopted in previous Attention models. The BRNN is effective for context information, but it is unable to extract high dimensional text features. Therefore, the CNN was introduced. The Attention model based on matrix transformation cannot characterize the features extracted by the CNN, a fully-connected neural network is used to improve the Attention model, and the NN-Attention is proposed. The recurrent neural network adopted was GRU, so as to speed up model training, the CSTSD dataset was used and TensorFlow was utilized for model construction. The experimental results show that the proposed model is able to realize automatic generation of text abstracts well in the CSTSD dataset.

Abstract:In order to improve the accuracy of WiFi-based indoor positioning and reduce the calculating time, this study proposes an indoor location algorithm combining Convolutional Neural Networks (CNN) with traditional fingerprint library. Based on the Received Signal Strength Indication (RSSI) data, the algorithm first uses the CNN model to predict the initial position of the measured point according to the real-time input data. Under the premise that the large-scale prediction position is guaranteed to be correct, the fingerprint points in the traditional fingerprint database are combined to determine the final prediction position with higher accuracy. The results show that the location accuracy of the error within 1 m is about 65%, the location accuracy of the error within 1.5 m is about 85%, and the error is stable under the premise that the timeliness is required.

Abstract:High-resolution remote sensing images contains rich geographic information. At present, the semantic segmentation model based on the traditional neural network cannot extract the features of small and medium-sized objects in remote sensing images, resulting in high segmentation error rate. This study proposes a method based on the connection of encoder and decoder structure features to improve the DeconvNet network model. The model can retain the spatial structure information by recording the location of the pool index and applying it to the upper pool when being encoded. During decoding, the model can effectively extract features by connecting the corresponding feature layer of encoder and decoder. During model training, the pre-training model designed can effectively expand the data to solve the problem of model over-fitting. The experimental results show that, based on the proper adjustment of optimizer, learning rate and loss function, the accuracy of remote sensing images semantic segmentation in the validation database is about 95% by using the extended dataset for training, which is significantly improved compared with the DeconvNet and UNet network models.

Abstract:With the aging population and more and more people living alone, a fall detection system based on dual cameras is proposed to reduce the damage caused by falls. Aiming at the ghost problem of Vibe algorithm in the process of moving target detection, this study combines the frame difference method to judge ghost area, which speeds up ghost elimination and avoids its interference. The minimum outer rectangle of the human body is used to mark the detected human body, and the height, aspect ratio, centroid, and Hu moment characteristics of the outer rectangle are obtained. The fall detection algorithm based on threshold analysis and Support Vector Machine (SVM) is used to judge whether or not the human body falls. In order to improve the detection rate of fall behavior, dual cameras are proposed for joint judgment. The experimental results show that the system can effectively distinguish fall from other daily behaviors, and the algorithm has high accuracy and real-time performance.

Abstract:At present, most PM_2.5 concentration prediction models only use time series data from a single station for concentration prediction, but do not take into account the regional correlation among air quality monitoring stations. This will lead to a certain one-sidedness of the prediction. In this paper, the KNN algorithm was used to select the relevant spatial factors in the area where the target site is located. Combined with the LSTM model, a KNN-LSTM PM_2.5 concentration prediction model based on spatiotemporal features was proposed. The simulation experiments were performed on pollutant data from 10 air quality monitoring stations in Harbin, and the KNN-LSTM model was also compared with other prediction models. The results show that the model compared with the BP neural network model, Mean Absolute Error (MAE), Mean Square Root Error (RMSE) decrease by 19.25% and 13.23% respectively; compared with the LSTM model, MAE and RMSE decreased by 4.29% and 6.99% respectively. It shows that the KNN-LSTM model proposed in this study can effectively improve the prediction accuracy of LSTM model.

Abstract:Lattice Quantum Chromodynamics (QCD) is a non-perturbative method for solving QCD based on the first principles. By simulating the interaction between gluon field and fermion field on super-cubic lattice, the calculated results are considered to be a reliable description of the phenomenon of strong interaction. Lattice calculation is of great significance to the study of QCD theory. However, the lattice QCD computing has a very large degree of freedom, which makes it difficult to improve the computational efficiency. Usually, the domain decomposition method is used to realize the scalability of parallel computing, but how to improve the efficiency of data parallel computing is still the core problem. In this work, taking Grid, a typical lattice QCD software, as an example, the data parallel computing pattern in lattice QCD computing is studied. Focusing on the complex tensor computing in lattice QCD and improving the efficiency of large-scale parallel computing, the theoretical analysis of data parallel computing features in lattice QCD method is carried out, and then the performance test and analysis are carried out for the specific data parallel computing methods such as SIMD and OpenMP of Grid software. Finally, the significance of data parallel computing pattern to the application of lattice QCD computing is explained.

Abstract:The traditional PV power generation prediction has the problem that the prediction accuracy is not high due to incomplete and inaccurate feature extraction of meteorological factors. In order to fully explore the influence of meteorological factors on PV output and effectively utilize the advantage of deep learning technology in non-linear fitting, this study proposes a short-term forecasting method for PV power that is based on the full mining of meteorological factors and is realized through the BiLSTM network. Based on outliers and standardized processing of the original data, KNN is used to fully explore the key factors affecting PV output among meteorological factors such as external temperature, humidity, and pressure. And then multivariate data sequences are reconstructed. On the basis of exploring the reasonable setting scheme of the hyper parameters such as the time steps of the input layer, the number of model layers, and the dimensions of each layer, a BiLSTM network model is built to realize the high-precision prediction of short-term power of PV power generation. Simulation results show that the proposed KNN-BiLSTM method has higher prediction accuracy than the classical methods such as KNN, DBN, BiLSTM, and PCA-LSTM.

Abstract:The quality of college graduates is directly related to the social reputation and development of colleges and universities. In order to accurately evaluate the quality of college graduates, based on the historical data of computer graduates in a university, this study uses an improved random forest algorithm to build a talent training quality assessment model. Before training classifiers, RF ranking method is used to measure the importance of features and select 75% of the features for dimension reduction, so as to improve the unbalanced phenomenon of training samples; through the training of base classifiers, the performance of each classifier is tested, and a single classifier is weighted according to the strength of performance, so as to reduce the impact of poor performance classifiers on the results. The practical results show that the algorithm improves the accuracy and precision of the quality assessment of personnel training, and can play a guiding role in personnel training in colleges and universities.

Abstract:With the popularity of online learning platform, a large number of learning behavior data are generated. How to use big data mining technology to analyze online learning behavior, to solve the problem that learners often face “resource overload” and “learning confusion” , better implementation of teaching decision-making, learning process optimization, personalized learning method recommendation, etc., has become a research focus. Based on the learning behavior data of Suzhou online education center, this work studies the common recommendation system model. Combined with the data characteristics of the platform, a collaborative filtering recommendation algorithm based on knowledge map is proposed. With this algorithm, the accuracy of the platform’s recommended resources is more than 90%, which effectively solves the problem of “learning lost” for students.

Abstract:Marine aquaculture is affected by a variety of diseases, and the differences in lesion characteristics are very suitable for image recognition. Based on the above requirements, this study designs a marine breeding disease diagnosis model based on VGG-16 convolutional neural network, and uses a stochastic gradient descent algorithm and overfitting prevention technology to improve the model. The experimental results show that this model is better than other traditional network models, and has high recognition accuracy, generalization ability, and robustness. It can accurately and quickly diagnose diseases with certain expansion and promotion value.

Abstract:The development of the instrument sharing platform has increased the utilization rate of instruments and equipment in various universities. However, during the use of the equipment, the fault detection of the equipment has not been improved. In view of the above problems, this study collected relevant data of medical imaging equipment, adopted the two-way feature selection method of PSO_RF for feature selection, then built a fault detection model based on LightGBM (Light Gradient Boosting Machine), and applied it to the fault detection of medical imaging equipment. Through the establishment of the standard evaluation system and the comparison of fault diagnosis results by different models, compared with the traditional machine learning algorithm, this model has a better performance in the accuracy rate, recall rate, F1 value and other evaluation indicators of fault detection, which has a positive role in accelerating the discovery of instrument fault points and improving the utilization rate of instruments.

Abstract:In order to generate pattern index on video stream in real time and quickly, an index authentication scheme based on Blockchain is proposed for event oriented real-time monitoring video query. By encrypting the secure channel between edge node and fog node, index data can be protected to improve the security of intelligent monitoring system. Firstly, by performing detection and tracking tasks on embedded edge devices, event oriented monitoring service extracts feature information by processing input frames. Then, real-time index service generates a unique index for each frame to prevent malicious modification of the image. Finally, the frame index is input into the Blockchain network and verified through the authentication mechanism based on decentralized smart contract Card. The experimental results show the feasibility and effectiveness of this scheme. The total cost is very low, which is suitable for the application of real-time monitoring video query.

Abstract:Sketches contain only simple lines and contours, which have completely different characteristics from natural images with rich colors and details. However, the current neural networks are mostly designed for natural images and cannot adapt to the sparseness of sketches. Aiming at this problem, this study proposes a sketch-based image retrieval method based on deformable convolution. First, the Berkeley edge detection algorithm is used to transform the natural image into edge map to eliminate domain differences. Then replace part of the standard convolution in the convolutional neural networks with deformable convolution, so that the network can fully focus on the outlines of the sketches. Finally, sketches and edge maps are sent to the network separately, and extract the fully connected layer features as feature descriptors for retrieval. Experimental results on the benchmark dataset Flickr15k show that the proposed method can effectively improve the accuracy of sketch-based image retrieval compared with existing methods.

Abstract:Nowadays, internet public opinion has a rapid spread and great influence, and topic detection plays an irreplaceable role in the supervision of public opinion. Aiming at the problems of incomplete feature extraction and high feature dimension in traditional methods, this study proposes LDA&&Word2Vec text representation model based on time decay factor, which combines the hidden subject features by LDA model with the semantic features by Word2Vec model, and adds time decay factor, which can reduce the dimension and improve the integrity of text features. At the same time, this study proposes a Single-Pass-SOM clustering model, which solves the problem of setting initial neurons in SOM model, and improves the accuracy of topic clustering. Experimental results show that the text representation model and text clustering method proposed in this study have better topic detection effect than traditional methods.

Abstract:National standard codes such as GB 18030 is the national standard of Chinese character coding in China, and UTF-8 is an international character encoding. In the internationalization, these coding methods exist simultaneously in Chinese information processing environment. In order to be compatible with the existing systems, such as document and protocol Chinese characters processing, the newly developed information system must convert the Chinese characters in the above form. In this study, the common Chinese character coding standards are introduced, and a Chinese character coding conversion program for lightweight applications of information system is described in detail, which supports the reuse of cross-operating system platforms.

Abstract:Kubernetes is Google’s leading container orchestration engine. Its resource scheduling algorithm consists of two processes: pre-selection and optimization. The pre-selection process needs to traverse all nodes, which is time-consuming, the improved scheduling algorithm proposes to directly optimize the number of nodes that meet the conditions without having to traverse all, which is expected to improve the scheduling efficiency. For the optimization process, only the CPU and memory usage applied by the pod itself are considered, and the resource utilization of the node itself is not considered. The improved resource scheduling comprehensively considers CPU, memory, network, IO indicators. The improved algorithm is verified through experiments, which can adapt to more complex internet application environments, thereby improving the load balancing efficiency of the cluster.

Abstract:Obstacle detection is the basis of autonomous movement of robot. In order to improve the efficiency and accuracy of obstacle detection, an obstacle detection method based on RGBD camera is proposed, which is mainly divided into two parts: obstacle identification, detection length and width. Under the premise of irregular shape, the obstacles through camera real-time image transmission to the data processing center, the improved frame difference minimum rectangle matching method and image processing method are used to determine the contour of the obstacle. Use of depth image and its threshold shows the relative position of the obstacle from the camera. The height and width of obstacles are detected by coordinate transformation method. The result shows that the error of detected object parameter is no more than 9%. Therefore, the improved frame difference method is of high accuracy in detecting the contour of obstacles, and the coordinate transformation method is of high speed. It can be proved that the obstacle detection design based on RGBD camera has sound detection effect.

Abstract:As the scale of systems continues to expand, the system structure also becomes very complex. The rule-based methods have been difficult to judge the composite faults under the interaction of multiple systems, and it is also hard to predict potential faults. Firstly, the study uses the ELK platform for centralized management of logs in complex scenarios of multi-business systems. Then, it sorts out the relationship between logs and various business systems, hosts, and processes in a complex system environment. Finally, we filter out the log files related to the failure in the system, and use these data in the deep learning framework TensorFlow to train the LSTM algorithm model, so as to realize the real-time fault prediction of the system.