Abstract:As the increased demand of modern manufacturing industry for product quality, production efficiency, and operation safety, video monitoring technology has been used more and more widely, which requires better precision and higher speed for video processing technology. Detecting moving object is the foundation for understanding and analyzing the video. Therefore, the algorithm for detecting moving object under industry background has been a hot research topic. Nevertheless, as an important component of this algorithm, the background features used in the moving object detection technologies have not received enough attention. Therefore, in this study, we demonstrate how to utilize visualization technology to extract the background features from the given pixel data and existing algorithm and then visualize the related features and data, which will help to obtain the relationships among these background features in a more straightforward way so that we can further explore them. The relationships among these background features can also be used for guiding algorithm design.

Abstract:The stopping scheme for passenger trains is important to the operation planning of trains, and the scheme affects the quality of passenger service and the transportation efficiency. This study established a multi-target programming model, aiming to minimize the total travel time of passengers and maximize the zone accessibility. In view of the traditional Particle Swarm Optimization (PSO) algorithm, which is inefficient and easy to fall into local optimum and cannot effectively handle the discrete problems when dealing with complex high dimensional problems, a new hybrid particle swarm algorithm is proposed based on the Quantum Genetic Algorithm (QGA). First, the algorithm adopted the construction of particle swarm algorithm, employing the idea of quantum bit coding, and using PSO algorithm velocity update mechanism to update the quantum revolving door. Since the algorithm combined the global exploration of QGA and intelligent system PSO populations, which not only improves the convergence speed of algorithm, but also increases the diversity of particle. Finally, the experiment on the ZDT function optimum and stopping scheme optimum problem shows that the proposed algorithm consistently provides faster convergence and precision.

Abstract:With the increase of data volume, the improvement of computer computing power, and the emergence of deep learning algorithm, artificial intelligence has been paid more and more attention. Taking American Core Journal Database Web of Science as research object, in which 6879 journal papers about artificial intelligence were included, adopting knowledge map of time and space and content knowledge map as main research methods, and applying the information visualization software CiteSpace, visualize and analyze literature big data from five aspects:cooperative countries, research institutions, references, keywords, and burst terms. It clarifies the research status and important literature in the field of artificial intelligence, and reveals the research hotspots and frontiers in the field of artificial intelligence. Finally, through the summary of the five visual analyses, the article gives an important reference in the field of artificial intelligence to select the direction of scientific research, to explore the forefront of science, to help science and technology decision, and so on.

Abstract:Increase of cyber-attacks is now becoming a serious problem. Among these attacks, malicious URL often plays an import role. It has been widely used to mount various cyber attacks including phishing, spamming, and malware. Detection of malicious URLs is critical to thwart these attacks. Numerous techniques are developed to detect malicious URLs and machine learning techniques have been explored with increasing attention in recent years. However, traditional machine learning methods require tedious work of features preprocessing and it is very time-consuming. In this study, we propose a detection method based solely on lexical features of URLs. First, we obtain the distributed representation of characters in URLs by training a 2-layer Neural Network (NN). Then we train the Convolutional NN (CNN) to classify feature images which are generated by mapping the URL to its distributed representation. In our experience, we obtained a reasonable accuracy of 97.3% and F1 of 91.8% using the real-world data set.

Abstract:Medical insurance fraud refers to the behavior of medical insurance fund or medical insurance coverage, which causes the loss of medical insurance fund through the method of deliberately fabricating and fictitious facts. Effective identification of health insurance fraud is of great significance to the rational use of health insurance funds. This study uses BP neural network to realize the active identification of health insurance fraud, and uses logistic regression analysis to improve the neural network model and reduce the interference of the weak factor to neural network identification. In addition, to deal with the scarce problem of fraudulent data, the model of neural network simulation function is used to train neural network. The empirical evidence shows that this method has better identification ability for health insurance fraud.

Abstract:The automatic control system of water and liquid fertilizer mixed irrigation, combination of open and closed loop, has been developed successfully based on wireless sensor network. In the closed-loop part of the system, the information of each greenhouse soil moisture is collected by end nodes and sended to coordination node after routing relay, the coordinator control water irrigation automatically depended on the information received. In the open-loop part of the system, the user instructions and the information of target tents, irrigation volume, the ratio of biogas liquid and water, etc. are received, then the system intermixes water and liquid fertilizer and conveys mixed liquor to target tents according to the instruction and the information automatically. The closed-loop part and the open-loop part of the system seamlessly integrate, and each part of the system can also work independently, and the system disperse business logic function to different nodes to realize distributed control. In addition to this, the system effectively improves the utilization of water and fertilizer and its cost is low, and has the high promotion and application value.

Abstract:In order to improve the design knowledge utilization and design efficiency in drilling assembly design, an intelligent design method based on ontology is proposed in this study. The method realizes the representation of the design knowledge of the drilling assembly and constructs the knowledge base of the design of the drilling assembly by constructing ontology model including borehole parameters, formation parameters, model knowledge, and drilling engineering base. In the method, the semantic mapping relations among ontologies are built. Thus an intelligent drilling assembly design pattern was formed and the organization, push and embedding of design knowledge are realized. Combining the case-based reasoning method, the intelligent design of drilling assembly is realized.

Abstract:In order to quantitatively analyze and qualitatively calculate the uncertainty and ambiguity of group users accurately, in this study, a qualitative similarity algorithm based on cloud model theory is designed to build a portrait model for group users. Firstly, the user is divided by the most widely recognized RFM mode-Recency, Frequency, and Monetary. Secondly, the user's behavior is transformed into the user's cloud model label through the cloud model transformation algorithm. The cloud model label is a quantitative representation of user's behavior. Then, the cloud model clustering algorithm is used to classify different types of customers which are the model of customer portrait. Finally, the model is used to guide commercial marketing activities.

Abstract:With the rapid development of open source technology, open source database has completed the replacement of commercial database in many business scenarios. The substitution of MySQL database for Oracle database is the most common. In the process of staged substitution, MySQL database and Oracle database often exist in the same set of business system, and MySQL database needs to synchronize data with Oracle database. At the same time, the deployment mode of multi node and high availability cluster in MySQL database further increases the difficulty of data synchroni-zation. By researching the method of data extraction of MySQL cluster by using Oracle GoldenGate software, we design a set of effective schemes to solve the problem of synchronizing data from MySQL cluster to Oracle database, and ensure the consistency and integrity of database in the process of synchronization.

Abstract:Aiming at the problem that the emergency treatment of high voltage power transmission line equipment failure is cumbersome, inefficient, and time-consuming, an Android mobile location and navigation system for equipment failure of power tower is designed. The overall architecture of the system and the client software structure are introduced firstly. Then this study focuses on the main interface design and the design of function modules, including loading power tower equipment module, fault location module, navigation module, information query module, and coordinate update module. By using the information of fault equipment, the system can accurately and quickly locate the position of power transmission line equipment failure. Furthermore, the system can directly navigate to the equipment failure of power transmission line and attain the information of faulty devices. As a result, it greatly reduces the finding time of equipment failure, also improves the speed and efficiency of emergency responsibility. So that it can save valuable time for the rapid recovery of power supply with strong practical significance.

Abstract:With the advent of WebRTC technology, the solution of real-time audio and video processing was provided. Its open standards provide a technical basis for the further research of real-time audio and video processing. Through the research of the core audio and video operation library of the WebRTC, we apply the technical solutions of WebRTC in the online interrogation module of the post-diagnosis medical service system, to improve the quality of the audio and video of online consultation service. Through analyzing the communication of the core audio and video of the WebRTC and depending on the real-time communication architecture of WebRTC protocol, we design the scheme of signaling agent and media agent, and finally realize the real-time communication of the post-diagnosis medical service system.

Abstract:With the rapid development of Internet, especially in recent years, the development of mobile Internet is increasingly rapid. Message push, an important way for mobile client to publish information and make communication, plays a significant role in mobile Internet. MQTT protocol is one of the implementation technologies of message push in Android system with low power consumption and high scalability. In addition, it can save Internet traffic, so it has been used in many applications. Meanwhile, as a distributed message queue, RocketMQ has great advantages in distributed deployment of servers. It has the characteristics of high performance, high reliability, and high real time ability. This paper introduces the MQTT protocol and the application of RocketMQ, and implements an MQTT protocol message push server based on RocketMQ through the combination of RocketMQ and Mosquitto and its distributed deployment.

Abstract:Feature point extraction and dimension measurement for human bodies has always been the key content of virtual garment fitting. Based on the human body image, this study realizes the extraction of human feature points and the size measurement by improving the ASM algorithm. Firstly, it calculates the distance between the two central points of face and body in the image, and matches them to the corresponding template, while changes the local template matching pattern in the traditional ASM algorithm. So the accuracy and efficiency of the initial model matching are improved. Then, it sets the feature point as the center and selects the less effective neighborhood points around the feature point for the object searching in the gray scale training model, which can solve the problem that the traditional ASM method takes long time and the feature points are easy mismatching. To solve that the unilateral fitting effect is better for the lower part of the human crotch, it uses the Mahalanobis distance formula, compares the gray scale of the specific matrix size neighborhood with the gray scale model, and combines with the human body shape distribution and symmetry feature to implement the feature point fitting process. Experimental results show that this method can adapt to the feature point extraction and size measurement of human body image in a complex background, and improve the extraction of human feature points and the accuracy of dimension measurement.

Abstract:This study proposes a saliency detection algorithm based on the fuzzy enhancement and feature points, using fuzzy enhancement and clustering segmentation to highlights the image object and internal classification. First, extract significant edge points and corner points, calculate the multiple features' means of those points, such as the brightness, color, and gradient features. Then, find all points which are belong to salient regions are closer to the means in the original image. By mathematical morphology to make sure the largest connected region, get salient regions finally. The experimental results show that the algorithm presented in this paper for saliency detection, can improve the accuracy and simplify the computation, the algorithm has an important role in the saliency detection, especially complex texture image.

Abstract:In the saliency detection algorithm, there are some problems in the detection of the manifold ranking, such as the over ideal of the background and the incomplete target detection. Aiming at these problems, this study incorporated background identification, BING feature estimation, and weight adjustment in traditional manifold rank algorithm, and a method was proposed based on background awareness. Firstly, through the adaptive color clustering of the boundary area and calculating the synthetic difference degree to get the real background seed point, the real background areas were sensed. Then the BING feature of the image was calculated and the saliency map information was combined to obtain the target position, so as to obtain the complete foreground seed point area. Next, by reconstructing the graph model of the foreground region and using the weighted k-shell decomposition method, we adjusted the connection weight between the nodes in the foreground region to obtain a clear target boundary. The experimental results show that the proposed algorithm is superior to other algorithms in terms of precision, recall, F-measure, and average MAE compared with some classical algorithms.

Abstract:The combination of multiple virtualization technologies can exploit the advantages of each platform and reduce the failure risks by the single one. In the heterogeneous environment, it is difficult to manage the virtual resources effectively, especially it might lead to serious waste because user's needs cannot be reasonable instantiated, and would suffer large cost and be difficult to integrate resources during machine migration. To solve the above problems, we propose an integrated model to manage the virtual machine resources across the heterogeneous platform, and bring a two level equilibrium allocation strategy based on Particle Swarm Optimization (PSO) to maximize the resources utilities for each platform. Under the extensive experiments in practical environment, the proposed method can improve the resource utilization and management efficiency.

Abstract:In order to better eliminate the targeted haze in image, an improved haze removal algorithm aiming to solve the shortage of original dark channel prior is proposed. First, the self-adaptive boundary is used to obtain the dark channel region block, and the interval of the atmospheric light intensity is estimated. Then, the transmittance repair method is improved. By including fault tolerance method, the improved algorithm can conveniently deal with the bright area which the original algorithm did not handle. The experimental results show that the improved algorithm can effectively remove the haze and bright area of the image.

Abstract:Compressed sensing is a sampling theory based on the sparsity of the signal. It has been widely used in the fields of compression imaging, medical imaging, radar imaging, astronomy, communication, and so on. The solution of the compressed sensing problem is essentially an optimization problem, on the basis of differential evolution algorithm. This study proposed an improved differential evolution algorithm, and the algorithm is applied to the solution of compressed sensing problem, and has achieved sound results.

Abstract:In the stage of spacecraft model design, the use of expensive licensed resources is extremely tight, the calculation efficiency is low. This study analyzed the shortcomings of the existing job dispatching mechanism in the high performance computing system that without considering the idle state of the high performance host. Based on the platform LSF, combined with the characteristics of spacecraft simulation analysis, a new resolution, i.e., rescheduling algorithm based on high CPU factor, which is used to distinguish the relative running speed of different machines, priority has been designed and implemented. The simulation results show that the algorithm can effectively improve the efficiency of the operation and shorten the occupancy time of licensed resources. The practical case shows that the algorithm has the possibility of popularization and application, which can improve the utilization of licensed resources and meet the needs of cost control for spacecraft simulation analysis and the efficacial utilization of resources.

Abstract:Aiming at the problem of low efficiency in obtaining training sample weights and solving the l₁ norm minimization, we proposed a face recognition algorithm WSRC_DALM algorithm, which was combined with Weighted Sparse Representation Classification (WSRC) and Dual Augmented Lagrangian Multiplier method (DALM). In the method, the Gaussian kernel function mainly was used to calculate the correlation between each training sample and the test sample, to obtain training samples with respect to the weight of the test sample. Then, the DALM algorithm was used to solve the l₁ norm minimization model, to achieve the test sample accurate reconstruction and classification. Finally, the proposed algorithm was validated by ORL and FEI datasets. In the ORL dataset, the recognition rate of the algorithm is 99%, compared with the classical SRC and WSRC algorithms, the recognition rate is improved by 7% and 4.8% respectively, and the computational efficiency is 20 times higher than WSRC algorithm. And in the FEI dataset, pose-varied face recognition rate is close to 92%. WSRC_DALM algorithm has obvious advantages in recognition accuracy and computational efficiency, and it has good robustness to large intraclass changes.

Abstract:In this study, we proposed a method to realize the embedding of DCT coefficients by selecting different priorities in order to solve the problem of non-perceivability, embedding capacity, and robustness in H.264/AVC (Advanced Video Coding) video hiding process, thus reducing the inter-frame drift distortion of information hiding algorithm. First, the secret information M is convolutionally encoded to obtain the encoded information, thereby improving the robustness of the video. Then, by analyzing the cause of the inter-frame drift distortion, the classification set of different Discrete Cosine Transform (DCT) coefficients priority order is calculated, and the data is embedded in the set of coefficients with small distortion to reduce inter-frame drift distortion. Finally, according to the pre-defined embedding rules, the DCT coefficients will be embedded in the set of selected luminance 4×4 blocks to increase the embedding capacity. Eventually, in the decoding side we can correctly extract the secret information while recovering the original information. The experimental results show that the algorithm proposed in this study can increase the video embedding capacity and improve the robustness under the premise of guaranteeing the video quality.

Abstract:The defects of overlapping community detection algorithm COPRA based on multi-label propagation include instability and pre-parameter limits, this study proposed a multi-label propagation algorithm for overlapping community detection based on LeaderRank and the node similarity. The algorithm uses the LeaderRank algorithm to sort the nodes in the network to determine the order of nodes updating. Then, re-design the label update strategy according to the similarity of nodes to improve the stability of the algorithm. The algorithm is applied to the artificial network and the real networks. The experimental results show that the proposed algorithm has high accuracy and stability for detecting overlapping communities.

Abstract:Aiming at the problems of large computation time and low classification time, this study presents an improved parameter estimation model for LDA by using the method of tensor decomposition, which can collect, classify, and mine massive network data. Using the method of moments, the LDA model calculation is transformed into low-dimensional tensor decomposition, and the parameters are transferred by decomposition and reflection. The large data platform Spark is used for distributed computation. The experimental results show that the model has been improved in terms of running time and perplexity, and the classification information display is more intuitive, which is more suitable for large-scale network data classification.

Abstract:As entering the market, Light Detection And Ranging (LiDAR) has been applied to the architectural planning, vegetation and water conservancy, and other industries since 1980s. Nevertheless, the application of 3D laser scanning LiDAR in terrain surveying and mapping is innovative, which has the advantages of high speed and accuracy. Its application can reduce the working intensity and time, improve the efficiency of surveying.This study gets the point cloud data as the source of data by using this technology for exploring the method of drawing contour indirectly. 3D douglas-peucker algorithm has been proved in order to sort points. Then, Isolines can be drawn by implementing the interface of arc objects. Finally, the number of points and grid resolution will be set in different levels in order to analyze their impact on drawing Isoline.

Abstract:Pedestrian detection is an important application of computer vision. However, it mostly uses the methods of low-level features. Deep learning, by combining the low-level features of pedestrians, can get more abstract representation of high-level features which makes the detection more robust. In this study, we propose a Faster Region-based Convolutional Neural Networks (RCNN)-based pedestrian detection method in which semantics is jointly considered. Firstly, we modify and fine-tune the Faster RCNN for fitting in the pedestrian dataset and for making it more capable of detecting small objects. Secondly, we establish connections between the pedestrian and its semantic attributes by spatial relationship, then fuse the pedestrian and its semantic attributes, and meanwhile adaptively adjust the confidence of the target pedestrian. The adaptive adjustment strategy, based on the connections between the pedestrian and its semantic attributes, realizes the fusion of the individual information. Extensive experiments and comparison show that the proposed approach in this study is of high accuracy, acceptable speed, and practical value. What is more, the semantic attributes can be used to count people or analyze the pedestrian's behavior.

Abstract:On the basis of analyzing the pattern data produced by the AutoCAD, CorelDRAW, and AI drawing software, the line turning point algorithm for pattern data, and reversing sewing and contracting sewing for non closed graph, and reinforcing sewing for the closed graph and the other customized algorithm were designed and implemented. The control algorithm of unilateral inflection point deceleration for the pattern data was studied and put forward. Test results show that the algorithm can efficiently and stably complete transformation from key point data of pattern pattern to pattern sewing point data. At present, this algorithm has been applied in pattern sewing machine software "intelligent software".

Abstract:The purpose of the study is to realize 3D reconstruction of human body based on Kinect. We use Kinect to scan and obtain the 3D human body data first. Depth data transformation algorithm converts 2D depth data to the global coordinate system. ICP registration algorithm achieves infrared camera gesture tracking to obtain the states of infrared camera. TSDF algorithm integrates the converted vertex set into the pre-divided voxel lattice. Projection mapping algorithm is used for visualization of the reconstruct results. Using Kinect and auxiliary equipment tripod, we can quickly and easily achieve the results of human body 3D reconstruction and output the model with 3D print. This research realizes the whole process of human body 3D reconstruction including scanning, processing, reconstruction, and 3D model output.

Abstract:GPS and wireless base staion positioning system integrated in terminal equipment has the disadvantages of obvious positioning errors and drifts under the condition of weak or no signals in some areas. This paper presents the hybrid location algolrithm based on particle filter combining AGPS, wireless base staion location, and Pedestrian Dead Reckoning (PDR). The principle of the hybrid location algolrithm is carefully described. Experimental results show that the terminal equipment can be accurately positioned even under the condition of weak or no signals of GPS.

Abstract:In order to solve the problem of time cost estimation in the development of geographically distributed software projects, the factors that influence development time and efficiency of geographically distributed development has be studied, and the actual situation of distributed software development is considered synthetically. The cost driving factors of COCOMOⅡ model are analyzed, screened, and newly increased. The COCOMOⅡ model is improved on this basis, so that the improved COCOMOⅡ model can reasonably estimate the time and cost of the distributed software development project. Finally, the development time and cost estimation of the improved COCOMOⅡ model are realized. The improved COCOMOⅡ model can better estimate the workload and the development progress of the distributed development.

Abstract:Since the rainfall and other uncertainties are difficult to effectively deal with in landside hazard prediction, as well as the density threshold in CFSFDP algorithm is required to be set manually and its low accuracy for large-scale data clustering, in order to improve the prediction accuracy, this study proposed an uncertain CFSFDP algorithm based on Grid and Merging clusters (uncertain GM-CFSFDP). Firstly, the E-ML distance formula based on uncertain data processing method is designed to effectively describe the uncertain factors of rainfall. Secondly, the idea of meshing is used to effectively encode the large-scale data by dividing it into multiple grid spaces. The average density of the mesh is calculated to establish the grid density threshold distribution model and obtain the grid density threshold dynamically. Finally, the hierarchical clustering idea is used to merge the higher association class and the uncertain GM-CFSFDP algorithm model is established. The experiments conducted in the Baota district of Yan'an show that the uncertain GM-CFSFDP clustering algorithm achieves a higher prediction accuracy and proves the feasibility and advancement of the algorithm in landslide hazard prediction.

Abstract:According to the extensive demand of two-dimensional (2D) bar code devices on the market, we studied the QR-code detection and recognition technology based on mobile DM642. This study proposes an algorithm based on position detection patterns, which have the nesting feature between the contour and the connection of area ratio. It locates the QR-code in order to solve the problem of skew and distortion of the image in embedded intelligent devices. Then, it precisely positions the four angular point coordinates of the QR-code by the method of linear approximation, and the QR-code is reconstructed by the principle of inverse perspective transformation and rotation correction. Finally, it obtains the whole 2D code "01" matrix by the grid sampling based on the reconstructed image, which is convenient for the decoding of QR-code. We transplanted the algorithm to the DM642 to run. It can effectively solve the image inclination and distortion problems and decoding effect is normal.

Abstract:In the process of investigation, exploration, and prevention about geologic hazard, a large number of heterogeneous data including text data is obtained. The method to storage the text data in file name search or large field is traditional, cannot meet rapidly retrieve and extract the useful information in the text data. It is an important problem faced by the geological hazard data storing and retrieving. In this study, based on the NoSQL, Chinese word segmentation, and Chinese keyword extraction technology, fast retrieval and statistics of any useful information are realized in geological hazard text data. It can provide strong support for the deep mining and fusion of hazard data.

Abstract:Estimating the level of informatization is a kind of complex, high technical content work, the problems mainly are the selection of indicators and the determination of weights in the informatization level measurement indicator system. Based on the existing informatization level measurement research, comprehensive geography and statistics, and other aspects of knowledge, through the word cloud analysis, coefficient of variation and correlation coefficient, factor analysis and objective weighting, and other technical means, a set of scientific and reasonable informatization level measure index system is established on the basis of amending the existing index system. The result is fitted and analyzed, which shows that the index system is reasonable and scientific.

Abstract:Because of low level of indoor medical information system, lack of remote monitoring, and insufficient medicine delivery research, the current system cannot meet the actual demand. In this study, we proposed an intelligent medicine delivery system framework based on cloud technology and Internet of Things (IoT); set up a new indoor medical and medicine robot delivery system based on embedded system, RFID, IEEE 802.11 communication protocol, and cloud technology. Then, we designed a remote mobile and PC terminal management system. Finally, we constructed the experimental and measured performances of the system and the mobile delivery robot. The experimental results show that the fastest control cycle of network nodes is 48 s per time, and the average period is 48.3 per time. The system delivers a medicine with an average time of 4.19 s. System upload cloud data, the fastest upload storage time is 2.1 s, the average cost is 5.21 s and the communication success rate is greater than 90%.

Abstract:An approach called "From-To Calculation Algorithm Based on Flow Graph" is presented. It is used to calculate the "from-position" and "to-position" of flow in complex piping network. By representing the piping network with graph topology, using the principle of maze backtracking, the direction of all the aspects can be determined in the piping system. A general component of the algorithm is developed using C# computer language based on generic parameter, interface technology and delegated observer mode. The algorithm is then applied to a mine drainage system. By taking the water pumps & joints as nodes, and pipes as edges, all flow "From-To" can be automatically calculated with "flow graph algorithm". Application results show that it can be an effective solution to calculate where a flow is from and where it will go.

Abstract:Text detection in the natural scenes is of great significance to the retrieval and management of large amounts of information such as video, images, and pictures. Depending on the complex background, low resolution and random distribution of the text detection in natural scenes, a scene text detection method was proposed, which combined the maximum stable extremal region algorithm and convolutional deep belief networks. In this method, candidate text region extracted from the maximally stable extremal region entered into the convolutional deep belief network for feature extraction. Then these features were classified by Softmax classifier. Experiments were carried out on ICDAR datasets and SVT datasets, and the experiment results show that the proposed method is helpful for improving the precision and recall rate of scene text detection.

Abstract:To solve the problem about low accuracy of traditional information recommendation method, this paper introduces the user portrait as the recommended basis, and proposes the recommendation method based on dynamic user portraits after further studying about the text classification and user behavior. By dynamically analyzing the user's historical data and predicting the user's interest trends, this method achieves dynamic recommendations. The off-line experiment improves that this method has some advantages in predicting user preference changes compared with the traditional label-based information recommendation, and it improves the recommendation accuracy.

Abstract:As the bandwidth of network become higher, it can provide a lot of web applications and services. But it also adds challenges to traditional system of data capture. In this study, we develop a high speed network packet capture software based on Intel Data Plane Development Kit (DPDK). This software can be used to capture the data of the high-speed campus network and provides technical support for network data package analysis. Lastly, we test this system and compare it with the packet capture system based on Libpcap. Experimental results show that this packet capture system can improve the performance of capturing high speed network data.