Abstract:The handoff process specified in 802.11 is time-consuming, and the strategy of choosing AP based on the received signal strength often fails to meet the mobile station’s bandwidth requirement. Therefore, 802.11 is unsuitable for industrial robots which have to rely on uninterrupted communication to work. This study proposes a fast handoff technology based on dynamic WiFi map that can provide bandwidth guarantee for mobile industrial robots. A prebuilt WiFi map is used to help the robots obtain their nearby Access Points (APs) without channel scanning, and at the same time, a server is used to collect the workloads of all APs periodically and provide robots with AP selection service that tries to meet both low handoff delay and bandwidth guarantee. This study implements the proposed fast handoff technology on a famous network simulation platform NS3 and compares it with some existing handoff schemes. The experimental results show that the proposed handoff scheme is superior to other handoff schemes in all cases, and can always choose the best APs with bandwidth satisfaction as long as there exists APs in the vicinity of the robot that have sufficient bandwidth.

Abstract:In machine learning, the label quality of the training samples seriously affects the final effect of the classification algorithms. Although the effect of a clean label is relatively good, it takes time and effort to collect and use. Therefore, in order to save costs and make the model adapt to the general situation, researchers gradually began to learn from ordinary data, that is, data with label noise. In recent years, although some works have been devoted to label noise, they lack comprehensive analysis. Based on this, this paper first introduces the label noise briefly and comprehensively, then analyzes the learning algorithms of tag noise in recent years explicitly and implicitly, and summarizes them. Finally, we look forward to the future research on label noise.

Abstract:To increase the users’ intelligibility of animated icons and enhance their cognitive efficiency in Human-Machine Interface (HMI), this study proposes an evaluation method of animated icons based on user satisfaction. Four animated icon forms including horizontal motion, vertical motion, rotation, and scale are redefined based on speed, path, amplitude, and direction. The psychophysical method is conducted obtain the upper threshold, the appropriate threshold and the lower threshold of the animated icons’ motion time, then the correlative models are built between the motion time and the motion amplitude of icons. The 7-point Likert-type scale is used to measure the appropriate motion amplitude of animated icon, motion time and amplitude of animated icons are compared and analyzed by ANOVA and LSD tests. The experimental results show that the most appropriate motion time was about 450 ms, the action time and the action amplitude is positively correlated amplitude and complexity of animated icons, even the size of smartphones is crucial to the design of animated icons. This method provides one quantitative model for the design of animated icons in HMI, the findings can serve as a useful therotical guideline when animated icons are researched or designed.

Abstract:In order to improve the reliability of software, software defect prediction has become an important research direction in the field of software engineering. Traditional software defect prediction methods mainly design static code metrics and use machine learning classifiers to predict the defect probability of the code. However, the static code metrics do not fully consider the semantic features hidden in the code. According to this situation, this study proposes a software defect prediction model based on convolutional neural network. First, extract the characterization vectors from the appropriate nodes in the abstract syntax tree of the source code, and construct a dictionary to map them to integer vectors to facilitate input to the convolutional neural network. Then, a convolutional neural network is designed based on GoogLeNet, and the ability of the convolutional neural network to deeply mine data is used to fully mine the grammatical and semantic features of the features. In addition, this model uses the method of random oversampling to deal with the imbalance of data, and uses the method dropout in the network to prevent the model from overfitting. Finally, the historical engineering database on Promise is used to test the model, and AUC and F1-measure are used as indicators to compare with the other three methods. The results show that the proposed model has a certain improvement in software defect prediction performance.

Abstract:In recent years, with the continuous improvement of environmental governance requirements, automobile exhaust becomes one of the main sources of pollution. The automobile industry is the focus of attention in China nowadays, and the driving conditions of automobiles are considered as reflections in the automobile industry. Therefore, the study of working conditions has become one of the urgently needed research projects. Based on various industry parameters of automobile driving conditions, this study develops a general method for reflecting automobile driving conditions in different regions of China. At the same time, we use the dimensionality reduction method of the de-noise encoder used in deep learning when reducing dimensionality of complex data, and has achieved good and practical experimental results. The data in this paper is derived from the automobile experiment in Jiading District, Shanghai. After processing the data, the construction of the general working condition map has been realized through different means and methods such as EMD, feature extraction, dynamic time planning, wavelet decomposition, etc., providing a reference for the construction of the general working conditions map of the city and the overall.

Abstract:Cloud storage is the future development direction of the storage business, and data security is the primary concern of cloud storage customers. The Ciphertext-Policy Attribute-Based Encryption (CP-ABE) algorithm allows the data owner to embed the access policy in the ciphertext and implement access control in conjunction with the key of data accessor, which is particularly appropriate for cloud storage environments. However, CP-ABE does not support time-related access control. This study proposes a CP-ABE scheme based on Time-Release Encryption (TRE). By incorporating a TRE mechanism in CP-ABE to achieve ciphertext sharing with time control, this scheme allows data owners to formulate a more flexible access strategy based on user attributes and access time. And then, we conduct security analysis to verify that this scheme can resist illegal access from users, cloud storage platforms and authorized institutions, as well as collusion attacks of illegal users. In addition, this scheme can also resist chosen-plaintext attack.

Abstract:Confidence measures represent the degree of match between speech data and models, and thus can be utilized to spot errors in voice command systems, improving their reliability. In recent years, systems based on identity vector (i-vector) and Probabilistic Linear Discriminant Analysis (PLDA) have been proven effective in the task of Speaker Verification (SV). This study proposes i-vector and PLDA as a confidence measure for voice command system without the need for acoustic or language models and demonstrates fair performance. Furthermore, in consideration of the deficiency of such i-vectors in modeling temporal information, this study proposes a fusion approach of such system with DTW, enhancing its time sequence discrimination ability.

Abstract:With the continuous accumulation of data, the data storage system has evolved from a single-node, small-scale file system to a distributed file system that supports massive data storage. How to provide simple, efficient and stable data uploading and downloading service for users has become an extensive research problem in distributed file system. To this end, we propose a server-side solution based on FTP protocol, which includes the use of buffer, coroutine, multi-process, and other technologies to improve the algorithms and optimize the transmission model, so that the data flow can be transferred rapidly. The proposed solution achieves high availability, high concurrency, and good scalability with the cooperation of Keepalived and LVS. Modifications to the FTP protocol makes the server support dynamic validation and compatible with multiple encoding formats. In the actual operation, taking storage system of China Science and Technology Cloud, i.e. iHarbor, as an example, the effectiveness, reliability, and performance of data transmission of the proposed solution are verified, which can fully meet the needs of users.

Abstract:With the rapid growth of meteorological data in Tianjin, CIMISS has insufficient storage and management for the mass data and its query efficiency is low in long-time-series data accessing. Therefore, based on the distributed storage technology and WebGIS technology, we design the Tianjin Integrated Meteorological Information Management and Sharing System. This system realizes intensive management, comprehensive display, unified service of Tianjin characteristic observation data, prediction products, intelligent prediction and shared data, it improves the access performance of Tianjin meteorological big data. This paper describes the design idea, data flow, key technology, main function and application of the system in detail.

Abstract:SpringBoot, as a new and popular framework at present, has unparalleled advantages in development, configuration, deployment and monitoring. Based on this framework, this study monitors and controls the key equipment of the distributed photovoltaic power generation system by separating the front and back ends. Unified collection, display, statistics, analysis and other functions of distributed photovoltaic power stations, and qualitative testing and quantitative analysis of system performance to ensure the safety, stability and reliability of the system, this project has been applied in actual projects, has achieved good results. Therefore, it has a certain social significance and promotion value.

Abstract:This study designs an indoor power transmission line thermal aging monitoring system in combination with the visible light sensors, infrared sensors and electromagnetic induction device. The electromagnetic induction device is used to detect the position of the power transmission cable in the wall. The infrared sensor uses a multi-point temperature measurement method to monitor the temperature of the wire in the wall or the wire in the distribution box. The visible light sensor is used for video monitoring and taking pictures of abnormal parts. If problems such as abnormal data are found, the device transmits abnormal data and photos to the host computer through the Ethernet communication module. Users can view the temperature of the current monitoring point, real-time video monitoring of the current monitoring point and set the temperature and electromagnetic force threshold through the mobile APP client or PC client. Once the monitored temperature and electromagnetic force data exceed the thresholds, the system will issue a line aging warning, the user can receive the mobile phone APP and PC client alarm push and picture information of abnormal parts, timely process the aging cables, avoid electrical fires caused by aging lines.

Abstract:The encounter between deep learning and big data technology has prompted face recognition technology to achieve a high level of accuracy. However, in actual application scenarios, especially in complex background, moving and natural face recognition, it has not yet achieved people’s satisfactory. Aiming at the problem of face recognition in attendance system, we propose using recursive minimum window algorithm to optimize the design of the face tracking algorithm in the M:N multi-face recognition scenario, and using multi angle sampling to improve the recognition accuracy and robustness. The proposed method is implemented and verified in the face attendance system. Achievement of multi-person synchronization within 3 seconds is achieved, and the user experience has been significantly improved.

Abstract:With the rapid development of road transportation industry in China, special transportation vehicles have increased significantly, which has brought great challenges to road travel and the safety of passengers’ lives and property. Based on massive special transportation vehicle driving data, this study proposes a multi-layer model of special transportation vehicle driving state evaluation system. Firstly, the data is processed for feature selection, outlier cleaning, and normalization. Then, the cluster analysis model and the dynamic threshold model are used to process vehicle driving data at the macro and micro layers, respectively. Finally, the results of cluster analysis and dynamic threshold analysis are combined to achieve a comprehensive evaluation of the vehicle’s driving status. The research results show that the multi-layer model proposed in this paper can make a more accurate assessment of the vehicle’s travel conditions and driving habits of vehicle drivers. It can provide reasonable scientific basis and data support for the management and supervision departments of special transportation vehicles and the vehicle transportation enterprises.

Abstract:In order to realize precise positioning with remote observation and control of the food delivery robot in the room, this study designs a indoor robot positioning information system based on the DW1000 module and the network module of ESP8266. The improved TWR algorithm is used in the system to avoid the problem of clock out of sync among the base stations and improve the clock offset error of the base stations. The simplified TOA algorithm makes the system locate quickly and accurately. At the same time, the wireless Ethernet module ESP8266 is used to transmit the position data and the status information to the remote server which is easy to view the real-time position state and control the robot movement. The system is an embedded wireless interconnection system, which can satisfy the indoor robot positioning and information transmission, and realize the remote control of the robot.

Abstract:In the event of a major natural disaster, communication equipment such as the base station in the area is vulnerable to damage, which is unable to meet the communication needs of the affected people. For this scenario, this study proposes a UAV-assisted file cooperative caching algorithm under the wireless mesh network architecture. Among them, the mesh router is the main cache device, the cache file block is used as the gene of cache strategy, and the genetic algorithm is used to implement file Cooperative Cache, which reduces the response delay of user request file, and achieves the purpose of emergency communication. UAV as the relay transmission node, takes the turning angle as the gene of trajectory planning, uses genetic algorithm to get the optimal relay position, assists the ground wireless mesh routing cache, and jointly provides file blocks for users. The simulation results show that the proposed algorithm can significantly reduce the average delay of users’ access to files and achieve the purpose of emergency communication.

Abstract:Aiming at the problem that the current Android malware detection method has insufficient ability to detect unknown applications, this study proposes an Android malware detection method based on the textCNN neural network model. This method uses a variety of trigger mechanisms to induce the potential malicious behavior of the application from different levels. For function calls at different levels, the specific hook technology is used to collect the application behavior. For the collected behavior logs, the fastText algorithm is used to extract word vectors. Finally, the textCNN model is used to detect and identify Android applications based on behavior logs. Experimental results show that the average accuracy of the method for detecting Android malicious applications reaches 93.3%, which verifies that the method has high effectiveness and correctness for detecting Android malwares.

Abstract:The existing image repair methods have some problems such as obvious trace, semantic discontinuity, unclear, etc. To solve these problems, this study proposes an image repair method based on a new encoder and context-aware loss. In this paper, the generative adversarial network is adopted as the basic network architecture. In order to fully learn the image features and get clearer repair results, SE-ResNet is introduced to extract the effective features of the image. At the same time, the joint context-aware loss training generating network is proposed to constrain the similarity of local features, so that the repaired image is closer to the original and more real and natural. Experiments on multiple public datasets in this paper prove that the proposed method can repair the damaged images better.

Abstract:With the increase of data dimension, the traditional clustering algorithm will have poor clustering performance. SubKMeans is a powerful subspace clustering algorithm, which aims to search the best subspace for K-Means algorithm and reduce the impact of high dimensions. However, the algorithm requires users to specify the number of clusters K value in advance, and sometimes it can not give accurate K value in actual use. In order to solve this problem, the pairwise constraint is introduced, which is combined with the silhouette coefficient. A SubKMeans algorithm for determining the number of clusters based on the pairwise constraint is proposed. The improved silhouette coefficient can evaluate the clustering performance more accurately, so that the K value can be determined. The experimental results proves the effectiveness of the proposed method.

Abstract:In recent years, with the rapid development of science and technology and the rise of deep learning, achieving image super-resolution reconstruction has become a hot research topic in the field of computer vision. However, the increase in network depth is easy to cause training difficulties, and the network cannot obtain accurate high-frequency information, resulting in poor image reconstruction. This study proposes an image super-resolution algorithm based on residual dense attention network to solve these problems. The algorithm mainly uses residual dense network, which accelerates the model convergence speed and reduces the gradient vanishing problem. The addition of attention mechanism makes the high-frequency effective information of the network have a larger weight and reduces the model calculation cost. Experiments show that the image super-resolution algorithm based on residual dense attention network greatly improves the model convergence speed, and the image detail recovery effect is satisfactory.

Abstract:The analysis of safety production accidents is of great significance to the improvement of emergency management ability. Based on the semantic analysis of safety production cases, Word2Vec embedding technology and clustering model are used, CBOW + negative sampling technology is used to realize word vector, and the data characteristics of safety production accident cases classification are combined, through semi-supervised learning based clustering model algorithm, according to the characteristics of the accident nature, an optimized initial clustering center algorithm is proposed, and K-means clustering algorithm is used to classify the text cases of safety accidents. The experimental results show that the proposed method can realize the classification of accident cases, and can be used for reference in the multi-dimensional analysis of accident.

Abstract:At present, the clustering method based on similarity is used to classify the wind power output scene, and the similarity is mostly measured by the Euclidean distance. Hence, the results reflect the difference of the amplitude of the time series curve, not the difference of the morphological characteristics and changing trend of the curve. This study proposes a method of wind power output scene division based on Gaussian mixture clustering, that is, the final attribution category is judged by the probability of belonging to a certain category. Firstly, the optimal numbers of GMM clustering and K-means clustering are determined according to BIC criterion, elbow rule and contour coefficient, respectively. Then, taking the actual wind power in a certain area as the research object, the typical scenes of wind power output in spring in this area are extracted, and the two clustering results are compared and analyzed to verify the effectiveness of this method. Finally, the typical scenes of wind power output in each season in this region are extracted by GMM clustering model.

Abstract:The pulse wave signal extracted by the non-contact heart rate detection method for a single feature region of a face video is susceptible to motion and light during the video acquisition process. In order to reduce the interference of motion artifacts and uneven illumination on pulse wave signals, a non-contact heart rate signal extraction method with multiple feature regions combined with fast independent component analysis is proposed in this study. Multi-feature regions are selected through the method of facial feature point algorithm combined with positioning regional center to ensure the stability of the feature regions in the video images. Fast independent component analysis is used to achieve the mutual compensation among the green channel blood volume change pulse signal of the images in the multiple feature regions, reducing the effect of uneven lighting. The experiment is performed on the DEAP data set published abroad. The experimental results show that the method in this study is superior to the existing methods based on independent component analysis and independent vector analysis.

Abstract:For the rectilinear parts cutting problem of low sheet utilization rate in practice, the concept of layout rectangle is introduced. The rectilinear parts cutting problem is decomposed into several optimization sub-problems. On this basis, the global optimal solution is constructed by solving the sub-problems based on dynamic programming. Experiment shows that compared with the traditional method of rectilinear parts cutting, the sheet utilization rate can be increased by 30%–50%. Compared with other typical algorithms, the sheet utilization rate is significantly improved and the scheme is more practical to layout.

Abstract:With the height of (super) high-rise buildings increasing, the crowd evacuation time in the building only through the stairwell will be incresed significantly. With the improvement of elevator safety technology, elevator-assisted evacuation technology in emergency is able to greatly raise the efficiency and safety for human. The Single Elevator Scheduling for Emergency Evacuation (S-ESEE) that considers the shortest evacuation time has been proved to be an NP-hard problem, but the limit on the number of elevators is not considered in the model. This study proposes a multi-objective model that minimizes the evacuation time and the number of round trips. The genetic algorithm is used to solve the model to avoid falling into the local optimal solution. To save the calculation time, some fixed values such as the number of people and the loss of elevator stop are calculated separately, and the time complexity of the algorithm is reduced by increasing the elevator stopping constraints. Numerical analysis results show that the two algorithms have little difference when the number of floors is small. However, as the number of evacuation floors increases, the algorithm in this study can obtain a better solution.

Abstract:In the dictionary learning algorithms, the model by the multi-vector representation can obtain better classification performance and more robustness than that by the single vector representation. In this study, we use the combined representation fused multiple vector representations and reasonable weighted logarithms sum schemes to improve the performance of the dictionary algorithm. Experiments on public face datasets verify that dictionary learning algorithms applied with proposed method has higher accuracy and robustness. It illustrates that the various potential appearances of observed objects generated by fully mining and utilizing the diversity of representations are beneficial to improve the performance of images classification.

Abstract:The traditional hand contour feature extraction can not deal with the effect of facial skin color, occlusion, lighting in the flight simulation environment. The traditional Fourier descriptor features are susceptible to background and hand posture changes, and have limited ability to describe gestures, etc. Hence, this study proposes a methed to improve the traditional hand segmentation and feature extraction methods. Firstly, skin color processing is performed on the collected data set, and then the 22 key points of the hand are detected in combination with the called hand key point model, and an eight-way seed filling algorithm is used for image segmentation. Then, the contours of the hand and key points are connected to the skeleton to extract the features of the Fourier descriptor algorithm. Finally, the Support Vector Machine (SVM) algorithm is used to train and recognize the extracted gesture feature data set. The experimental results show that the method in this study has good hand segmentation, feature extraction is not easily affected by changes in the background and hand posture. Hence, it can well cope with interference in a complex background in a flight simulation environment, with the recognition accuracy reaching 98%. The research presented in this paper has a certain role in improving the traditional gesture recognition algorithm and has very important application value in the field of hand interaction technology.

Abstract:Based on the rutting depth index and driving quality index in the pavement evaluation index, the pavement damage was evaluated in this study. The association rules were used to mine the degree of association between influencing factors such as environment, traffic, and road surface and road surface conditions. Aiming at the shortcomings of the complexity and time-consuming of the association rule Apriori algorithm, an improved Apriori algorithm that does not generate candidate sets to generate frequent sets was proposed. The experiments show that the improved Apriori algorithm can effectively improve the speed and performance. The improved Apriori algorithm was used to analyze the strong association rules between evaluation indexes and influencing factors, and the main causes of pavement damage in different environments were obtained. The conclusion of this paper can provide scientific and reliable support for the pavement maintenance, reasonable maintenance suggestions, and data support for the pavement maintenance department.

Abstract:Cross-domain object detection is a new research direction, which aims to solve the problem of generalization from training set to test set. In the existing methods, using image style transfer and train the model on the converted data set is an effective method. However, this method has the problems of not end-to-end training, low efficiency, and tedious process. Therefore, we propose a new cross domain target detection algorithm based on image style migration, which can combine image style migration and target detection to carry out end-to-end training, and greatly simplify the training process. The results on several common datasets show the validity of the model.

Abstract:Accurate prediction of traffic flow is of great significance for safeguarding public safety and solving traffic congestion, and plays an important role in urban traffic planning, traffic management, and traffic control. Traffic forecasting is one of the challenging topics in recent years because it is restricted to urban road networks and changes with time, and there are spatial dependence and time dependence. In order to capture both spatial and temporal dependencies, a new neural network is proposed: A space-time map convolutional network based on the attention mechanism (A-TGCN). The TGCN network model is used to capture the dynamic spatiotemporal characteristics and correlations in traffic data, and an attention mechanism is used to enhance the information of key nodes in each A-TGCN layers. The experimental results on two sets of data show that A-TGCN has a good performance in terms of accuracy and interpretability.

Abstract:With the vigorous development of the logistics industry, the development of smart warehousing logistics can greatly reduce logistics costs and accelerate the development of the industry. This study proposes an AGV vehicle system collision avoidance path planning in smart storage. First, the grid method with time windows is used to simulate the working environment of the AGV’s manufacturing workshop, and an improved ant colony algorithm is proposed. By improving the probability conversion formula and the pheromone update rule, the simulation results verify that the algorithm can solve the obstacle avoidance path planning problem of multiple AGVs, and then realize the intelligent warehouse logistics.

Abstract:In recent years, the convolutional neural network model is often used in the research of text emotion classification. However, most of researches ignore the emotional information carried by the text feature words themselves and the wrong segmentation of Chinese text. Aiming at this problem, a Dual-channel Convolutional Neural Network sentiment classification model fused with Sentiment Feature (SFD-CNN) is proposed. In the model, one channel is used to construct the semantic vector matrix of emotional features to get more emotional type information, and another channel is used to construct the text word vector matrix to reduce the impact of segmentation errors. The experimental results show that the accuracy of SFD-CNN model is as high as 92.94%, which is better than that of the unmodified model.

Abstract:It is difficult to recognize microexpression because of its small range and short duration. To solve this problem, a micro expression recognition method based on 3D Convolutional neural network (C3D) and optical flow method is proposed. We first extract a series of optical flow images with dynamic features from the microexpression video by optical flow method, then input the obtained series of optical flow images with the original gray-scale image sequences into the C3D network, and then extract the features of micro expression in the time and space domain by C3D. Simulation experiments on the open data set CASMEⅡ show that the recognition accuracy of the proposed method is 67.53%, which is better than the existing methods.

Abstract:Non-contact Heart Rate (HR) detection can be achieved by remote PhotoPlethysmoGraphy (rPPG), which has attracted more and more attention. However, in practical applications, the rPPG signal is very subtle and easily overwhelmed by noise, which makes it difficult to accurately estimate the HR using existing rPPG-based HR detection methods. In view of the above problems, this paper proposes a non-contact heart rate detection method that enhances rPPG signal and suppresses noise. In this method, the chromaticity information in the normal HR distribution band is first amplified by Euler color amplification technology to avoid the rPPG signal being too small and being overwhelmed by noise; then use face detection and tracking technology to select the appropriate skin of interest Region; then extract the amplified chrominance information in the region of interest, and use the blind source separation method and correlation analysis to separate the rPPG signal; finally, the rPPG signal is time-domain filtered and power spectral density analysis to estimate the HR value. Multiple experiments show that the proposed method has higher HR estimation accuracy than previous methods.

Abstract:In view of the problem that charging facilities in intelligent park cannot meet the development demand of electric vehicles, a modeling and analysis method suitable for orderly shared charging demand of electric vehicles in intelligent park is proposed. Charging plans are provided by analyzing charging data, and orderly charging strategies are formulated in the energy management system of the park. Power spectral density estimation method using statistical analysis of single EV charging current, and complete EV on-line identification in artificial intelligence network classification work, based on the correlation between insertion time, energy, and working day statistical analysis each EV charging habits, the grid side current is measured using to predict the charging demand, based on kernel density estimate charging demand statistics model is set up. The validity and applicability of the model are verified by the actual data collected from a charging facility in a residential area, which can provide help for EV charging in the intelligent park.

Abstract:In order to solve the problem that the traditional model and the machine learning model have poor performance in predicting the periodic time series, this study proposes the SSA-LightGBM prediction model which takes the Hancheng expressway traffic volume survey data as the data set. First, the Passenger Car Unit (PCU) of Hancheng expressway data is calculated. Second, the singular spectral analysis is applied on the PCU data to obtain periodic and random terms, the periodic term is reconstructed, and then the LightGBM is used to predict the random term. Finally, the predicted random term and the periodic extension signal are superimposed to obtain the prediction data of final expressway PCU. At the same time, compared with the prediction results of XGBoost and LightGBM model, the prediction results of SSA-LightGBM are closest to the true values, and MAE, RMSE and R² are better. This result has a good guiding significance for the research of the forecast on future change trend of our country’s expressway volume survey data, and provides a good reference value for the renovation and maintenance of our country’s expressways.

Abstract:Mainly based on the image sequence, this study reconstructs the movement track of table tennis, and analyzes the movement form of table tennis. Firstly, the collected image is stereo corrected, and the center coordinates of the table tennis in the sequence image are extracted by color recognition and improved Hough circle detection algorithm; then, the feature points are matched on the time series according to the difference of the feature point coordinates of the front and back frame images; finally, the matching feature points are 3D reconstructed by triangulation, and the speed and acceleration of the table tennis at different times are calculated, the 3D motion reconstruction of dynamic object is realized. The experimental results show that the 3D motion reconstruction method improves the accuracy of feature extraction, effectively realizes the matching on time series, and obtains the 3D motion data of the object.

Abstract:In order to find the complex rules in the data, avoid the blindness of kernel function selection and local optimal nonlinear optimization problems, this study proposes an improved gray wolf algorithm to optimize the multi-core support vector regression machine algorithm. Firstly, a multi-core SVM oil production speed prediction model is constructed based on the global kernel function and the local kernel function. Secondly, the gray wolf optimization algorithm is improved based on the cloud model and the quadratic interpolation algorithm to optimize the selection of the weights and parameters of the kernel function. Finally, the influencing factors set of oil production speed is determined by the grey correlation analysis theory and used as the multi-core SVM prediction model. Compared with 6 kinds of prediction methods of oil production rate, the proposed method has the advantages of better global optimization ability and higher prediction rate.

Abstract:Authentication key agreement is vital for secure communication on the public network, it can make communication in a malicious attacker current safely set shared session key. Certificate-Based Cryptography (CBC) to solve the certificate revocation problem in traditional public key cryptosystems, the problem of key escrow in identity-based cryptosystem and no certificate cryptosystem in the security channel problems is established. The existing certificate-based authentication key agreement scheme is mostly adopted the expensive bilinear pairing, not suitable for calculation with limited resources of mobile devices. In this study, we design a lightweight AKA protocol based on the certificate, the protocol uses pseudonym technology to realize user anonymity, and provides forward confidentiality, man-in-the-middle attack resistance, replay attack and other security analysis. Compared with the previous certificate -based AKA protocol, this protocol has obvious advantages in computing efficiency.

Abstract:Aiming at the defects of porous materials such as blockage and angle gap, a method of surface defect detection for porous materials based on machine vision is designed in this study. By means of effective segmentation of target area, ambiguity detection, morphological treatment and analysis, the rapid location and feature analysis of surface defect for porous materials are realized. The experimental results show that the accuracy and efficiency of the algorithm can meet the real-time detection requirements of industrial production.

Abstract:The problem of the absence of attribute data often occurs in soil analysis and research. To improve the reliability of the research results, it is necessary to study the imputation methods for soil attribute missing data. In this study, a variety of imputation methods have been evaluated to interpolate the soil attribute missing data from the perspective of data mining. Using soil attribute pH as an interpolation object, the Soil Nutrient Database of China’s Major Ecosystems is used as the source of physical and chemical soil attribute data. We evaluate the performance of each method on the dataset of different missing rates in terms of model fitting and imputation error. The result shows that it is feasible to impute soil attribute pH missing data using the optimal parameter K-Nearest Neighbor (KNN) and random forest than other methods, such as multivariable regression, support vector machine, and neural network. The mean value of MAE、RMSE and R² of the imputed missing data pH of KNN and random forest on the dataset with different missing rates are 0.132 and 0.131, 0. 174 and 0.178, 0.775 and 0.765, respectively.

Abstract:Aiming at special situation under COVID-19 epidemic environment, big data technology is used to process the power data of different data structures at the very beginning. Firstly, the data of users’ electrical behavior is divided into internal data and external data. Secondly, the behavior of electricity consumption is analyzed by improving K-means clustering algorithm, which can improve the efficiency of the traditional K-means algorithm. Finally, the improved K-means algorithm clustering is used to build the power consumption behavior model, so as to realize the results of the users’ power consumption behavior analysis model. This study helps the state grid corporation of China to achieve the purpose of intelligent distribution of electricity, and gives the general policy tendency, which improves the regulatory capacity of all government departments to help manage national emergencies.