Abstract:At present, scene text detection based on deep learning has achieved good performance in complex background. However, it is difficult to precisely detect text with small scale. To solve this problem, this study proposes a deep neural network based on feature fusion, and a new neural network with senior semantic is constructed by combining the high-level feature and low-level feature of traditional deep neural network. Strong semantic information of the high layer network is utilized to improve the overall performance of the neural network, and the feature fusion network directly predicts text with multiple scales through multiple output layers. Experimental results on ICDAR2011 and ICDAR2013 datasets show that proposed method is significantly effective in detecting small scale text. Meanwhile, the proposed method has high accuracy and robustness in scene text detection, and the F-measure achieves 0.83 on both datasets.

Abstract:Locating human eye accurately and separating iris and eyelid is of great significance to biometric identification technology, such as iris recognition and face recognition. However, segmentation for the human eye under non-ideal environment is a very challenging job due to lower resolution of eye image, noises like superimposed occlusions of reflections, eyelashes, or shadows. In this study, in view of existing problems for unsheltered human eye image with small pose, we adopted circular Hough transform and morphological algorithm to improve the location of human eye in low resolution. Firstly, this method segments region of interest in face picture with existing face alignment method, and then makes use of bilinear interpolation to remove reflection in image processing module. Secondly, according to the gray distribution of human eye image, we utilize the constrained circular Hough transform to locate iris. Then, combined the global dynamic threshold method, local adaptive threshold method and morphological algorithm are used to locate upper and lower eyelids. After that, fit the eyelid using the least square method so as to cut apart the iris, sclera, upper and lower eyelids in human eye image. Finally, the proposed method was tested on low resolution face image in the wild and UBIRIS V1.0 iris images database released by University of Beira, Portugal. Experimental results demonstrate that the proposed algorithm has strong robustness for human eye location and segmentation in laboratory environment or for low resolution images in the wild.

Abstract:With the rapid development of the Internet of Things (IoT), a huge number of embedded devices are widely used in modern life. Security and privacy have become an important challenge in IoT. Interconnected devices in IoT constitute swarm network. The state trust of swarm devices, or swarm attestation is the key issue for the security of IoT. Traditional remote attestation schemes mainly concentrates on single prover scenario and are lack of large-scale device swarm attestation technology. Targeting at the low efficiency and vulnerability to collusion attack in existing device swarm attestation, in this study, we propose an efficient swarm attestation scheme based on devices grouping. This scheme groups homogeneous devices and sets up a management node in each group to verify the rest of nodes in the same group. When the swarm executing the swarm attestation scheme, the remote verifier only need to verify the management node, because each management node has known its own group's trust state. By this way, this scheme improve the efficiency. It also has high security and is able to resist collusion attack. The prototype system experiment results show that the more homogeneous devices and the less management nodes, the efficiency of the proposed scheme is higher.

Abstract:The N-gram model is one of the most commonly used language models in natural language processing and is widely used in many tasks such as speech recognition, handwriting recognition, spelling correction, machine translation and search engines. However, the N-gram model often presents zero-probability problems in training and application, resulting in failure to obtain a good language model. As a result, smoothing methods such as Laplace smoothing, Katz back-off, and Kneser-Ney smoothing appeared. After introducing the basic principles of these smoothing methods, we use the perplexity as a metric to compare the language models trained based on these types of smoothing methods.

Abstract:Microsatellites are short repetitive sequences widespread in eukaryotic genomes. Microsatellite instability, the gain or loss of repeat units from repetitive DNA tracts caused by MisMatch Repair Deficiency (MMRD), is significant for tumor early diagnosis and prognosis. In current clinical practice, microsatellite instability detection is performed by experimental methods like MSI-PCR and MMR-IHC. With the development of next-generation sequencing technology, a number of MSI detection softwares utilizing high-throughput sequencing data have been developed. In this paper, we provide an overview on existing MSI detection methods, including experimental methods and computational methods, as well as their strengths and limits.

Abstract:Compared with traditional Enterprise Data Warehouse (EDW) solution having vendor lock-in risk on both hardware and software, Massive Parallel Processing (MPP) database and Hadoop/Spark ecosystem based on open hardware platform, like X86, have advantages on both scalability and low cost. And with the mature of these technologies, they become integral parts of enterprise-class big data platform. Based on how to deal with the challenges posed by "big data", this study proposes a fusion big data architecture suitable for data analysis and processing of large commercial banks, and has achieved the success of application practice.

Abstract:For the existing business analysis system of the massive data storage, massive data processing, and business development needs, we designed and built the enterprise business analysis system based on big data cloud analysis platform. This study focuses on the construction of the system technology route, the overall framework of the system to achieve and operational results. The results show that the system can improve the efficiency of data utilization and can provide better data support for the operation of the enterprise.

Abstract:The planning for actions of characters plays a prominent role in displaying the animation theme. However, the automatic generation system of mobile phones 3D animation lacks of the planning for interactive actions of characters. This paper presents an interactive planning method based on the three-layer action identification and Kinesiology. This study mainly looked at two aspects:the qualitative planning and the quantitative calculation. First of all, semantic web technology was used to set up an interactive action ontology base. In the qualitative planning stage, the theme and template of information extraction were used to form interactive actions corresponding to the content of the short messages. Then, in the quantitative calculation stage, determine the physical posture of both parties performing actions and to plan the temporal and spatial relationship between the two parties of interactions. The experimental results show that the proposed scheme can make the characters in the animation perform more complex interactive actions.

Abstract:The microservice architecture facilitates the service decoupling of application services and the separation of the technology stack. However, more microservices also increase the frequency of stateless service invocation across processes. How to ensure secure service access control between stateless services while ensuring service performance is a key issue for the microservices security architecture. In this study, we design a flexible microservice security access control framework. Combining the features of microservice API gateway, the lightweight microservice token construction mechanism and the flexible adaptation of microservices security control strategy, we improve the flexible security control ability of microservice. After the experimental analysis, the cost is smaller, and the validity of the framework and the method is verified in the actual project.

Abstract:Qinghai Lake is the largest inland lake and saltwater lake in China. It is the habitat for migratory birds and has important ecological research value. In order to effectively manage and apply the ecological monitoring data collected in Qinghai Lake, this study designed and implemented the Qinghai Lake ecological hydrological monitoring data visualization platform. The platform uses Spring MVC as a background development framework. The database uses the open source timing database InfluxDB and uses InfluxDB's data retention strategy and continuous queries to aggregate data. The visualization function is implemented by Echarts, an open-source JavaScript tool. It can visualize the Qinghai Lake's ecological data such as atmospheric environment data, hydrology data, and soil data.

Abstract:The emergence of semantic sensor Web provides a way for realizing data interoperability, information sharing, and knowledge fusion among the different sensor systems in the Internet of Things (IoT). In this study, taking some sensors and corresponding data acquisition instruments as examples, the ontology based on the Semantic Sensor Network (SSN) proposed by World Wide Web Consortium is designed for providing the meaning background information and the knowledge organization about the sensor systems on the Web. Automatic classification management is realized by using this ontology, and the collection between the sensors and the instruments can be recommended according to the reasoning rules. The efficiency and the reliability of system are improved, and the domain knowledge can be also combined into the system. Finally, the reasoning rules are tested with the specific sensor example and the results meet the application requirements.

Abstract:With the extensive use of CAD drawing, drawing safety is widely concerned. This study mainly focuses on the operation safety of engineering drawings under AutoCAD environment. A safety management system for engineering drawing is developed with ObjectARX technology. In this system, ARX reactor technology is employed to monitor and control AutoCAD printing and save operations. API HOOK functions are used to intercept the clipboard function in order to change its execution results, and to control the copy and paste operations. In addition, the PrtSc key and Screenshot operations are controlled. The test results show that the system can monitor and manage the user operation, and effectively ensure the integrity of the electronic drawings, at the same time, it does not affect the user's usage habits.

Abstract:In order to meet the growing demands for high concurrency, quick response, dynamical scaling, and maintainability in larger-scale internet applications, in this research work, a distributed cluster cache system based on Redis 4.0 has been implemented, which the data can be distributed and scaled into different nodes so the system's linear scalability, load balancing, concurrency, data throughput, and responsiveness can be optimized. The open source tool called CacheCloud is integrated into it so the cluster can be efficiently managed and monitored in real time. The results show that the system reaches high performance and response time of using Query Per Second (QPS) on Redis Cluster is much faster than that on Codis after 10 000 concurrent accesses.

Abstract:The distributed and complex trends suggest the new requirements for customization, extensibility and agility of the enterprise architecture modelling methods. In this study, three kinds of mainstream enterprise architecture modelling methods Zachman, FEA, and TOGAF are introduced and compared, and the high-level ArchiMate modelling standard with the existing low-level modelling standard UML are also compared and evaluated. Through a real case study, this paper demonstrates the application of designing enterprise architecture using ArchiMate, and how to customize ArchiMate in real environment.

Abstract:In order to achieve the various parameters in the process of underground coal gasification of real-time monitoring, data analysis, and the realization of accurate control of the production process of all key variables, this study designs the underground coal gasification process measurement and control system based on PLC. This paper first introduces the underground gasification process and testing requirements analysis, secondly presents the design of overall scheme and hardware and software, finally elaborates the database and mathematical model used in the applications. The mathematical model of measurement and control system consists of three layers of topology:device layer, control layer, and monitoring layer. The equipment layer includes the basic hardware devices. The control layer includes two Siemens S7-300 series PLC station, constituting the double CPU soft redundancy system, to ensure the stable air supply in ignition lane when one PLC station fails. Monitoring layer uses the configuration KingView and Navicat Premium software. Above design realizes the real-time monitoring of data display, video image real-time monitoring, historical data storage, and so on. This project has been successful run for nearly a year in Shanjiaoshu Mine affiliated with Guizhou Panjiang Reined Coal Co. Ltd..

Abstract:As part of the Android security model, SEAndroid is critical to assure the security of operating systems. In this study, we propose an approach to analyze SEAndroid policies based on capability dependency graph. Capability dependency graph describes attacker's potential capabilities and the dependency relationships among these capabilities. It also describes the configuration of SEAndroid policies. We collect some security related system facts, and encode the collected data to Prolog predicates. We adopt logic programming to automatically compute a capability dependency graph with driving rules. We enumerate all the attack paths from initial nodes to goal nodes in the capability dependency graph, and categorize the attack paths into attack patterns. We apply our approach to analyze and compare some different versions of Android. We find that with the upgrade of the Android version, the SEAndroid security policy has also been updated. The new SEAndroid provides a stronger constraint and protection for the system, and a experimental attack pattern has been verified in the actual system.

Abstract:With the advent and popularization of Internet mass media, massive data have been generated in forms of texts, images, videos, etc. This poses a serious challenge for the review of related content, especially the security review of the image data. At present, the safety analysis of image data is not mature and thus criminals often hack websites and tamper with the images on the websites, which poses significant threats to network security. Targeted at this practical application, this study designs and implements a web image data security analysis system, which is composed of two major modules:(1) illegal image detection engine module based on deep learning algorithm; (2) image tamper-proof module based on event trigger technology and plug-in polling technology. The system can quickly review whether the image data content is legitimate and automatically monitor whether the image data has been tampered with.

Abstract:The submarine cable integrated online monitoring system is mostly used in major offshore projects such as offshore oil platforms and offshore wind farms. In order to realize the remote data release function of the submarine cable online monitoring system and make data transmission more efficient and reliable, this study deeply analyzes the structure model and communication process of the IEC60870-5-104 protocol. On this basis, we propose a set of submarine cable online monitoring data based on the protocol and explain the design ideas and level model of the system. Simultaneously, we propose the corresponding solutions for the two key issues of dynamic configuration and mass data transmission. Finally, the feasibility of the scheme is verified by an application example of a marine engineering project.

Abstract:Traffic big data is the most basic condition for solving urban traffic problems. It is an important guarantee for formulating macro-city traffic development strategy and construction planning. And it is also an important guarantee for carrying out micro-road traffic management and control. In view of the characteristics of intelligent transportation system such as fast data generation, high real-time performance and large amount of data, this paper constructs a real-time traffic data processing platform based on the combination of Spark Streaming and Apache Kafka to process the data collected by dual base stations. Using time window mechanism to get data from Kafka, and process the data to the database according to the rules. In this study, the system architecture and internal structure of the platform are introduced in detail, and the real-time processing capability of the system is verified through experiments, which can be applied under large-scale and high-concurrency data flow.

Abstract:With the rapid development of high-speed railway, the research of the data secure transmission of high-speed railway based on cloud computing and real-time online analysis have become the hot topics. Aiming at the problem of the data secure authentication in the process of the data transmission at high-speed rail, a data secure transmission protocol for high-speed rail sensor system based on SM2 signature algorithm is designed. The traffic state data collected by the sensor network are transmitted to the cloud service platform by the security protocol based on SM2 signature algorithm, and the secure interaction between the high-speed rail traveling data recorder and the cloud service platform is realized. Therefore, the reliability and integrity of data transmission are improved. Finally, the experiment and simulation of secure transmission protocol are carried out, and the result shows that the protocol has high efficiency and security in the process of high-speed railway data transmission.

Abstract:Metric, also called distance function, is a special function in metric space that satisfies certain conditions. It is generally used to reflect some important distance relationships between data examples. Since distance has a great influence on various classification and clustering problems, metric learning has an important influence on these machine learning problems. Existing metric learning algorithms for classification problems are vulnerable to noise, the classification accuracy is not stable and tends to fluctuate. To solve this problem, this paper presents a robust metric learning algorithm based on maximum correntropy criterion. The core of maximum correntropy criterion is Gaussian kernel function, which is introduced into metric learning in this study. We construct a loss function with Gaussian kernel function and optimize the objective function using gradient descent method. The output metric matrix is computed through repeatedly testing and adjusting the parameters. The metric matrix learned through this method will have better robustness and will effectively improve the classification accuracy when dealing with various classification problems affected by noise. This study performs validation experiments on some popular machine learning datasets (UCI) and face datasets.

Abstract:There are some anomalies before the earthquake, especially the large earthquake. However, such abnormal information is too difficult to identify. Therefore, we cannot make full use of the abnormal information to predict the occurrence time of the earthquake in order to reduce the impact of the earthquake. To solve this problem, an anomaly mining method before earthquake based on the quantum walk algorithm is proposed to extract seismic Outgoing Long-wave Radiation (OLR) anomalies before the Wenchuan earthquake and the Lushan earthquake. Then, calculate the P value, anomaly value CD before and after the earthquake. Through statistical analysis method, the relationship between OLR anomalies and earthquake is explored. What is more, the algorithm is extended to the 8.0 magnitude and above earthquakes in the nearly last ten years. Through experiments, the effectiveness of the algorithm is verified. The experimental results show that the algorithm can effectively reflect the anomalies before and after the earthquake, and the larger the earthquake is, the more obvious anomaly is. Therefore, this algorithm is suitable for pre-earthquake anomaly excavation.

Abstract:Aiming at the problems such as false matching points and large volume remote sensing image registration, a high-resolution remote sensing images registration algorithm based on coarse and fine registration is proposed. Firstly, the high scale space feature points were extracted after down sampling the images to execute the coarse registration. Secondly, the initial set of feature points was extracted using the Scale Invariant Feature Transform (SIFT) algorithm for each block after using image blocking strategy. Furthermore, feature points were used to obtain the Delaunay triangulation, and then calculated the similarity between blocks of both images to select pairs of triangles which the similarity greater than threshold. Finally, the fine registration was achieved by precise feature points. The proposed algorithm can reduce the number of feature points, and it can eliminate more false matching points to increase the correct feature point matching rate. The experimental results indicate that the proposed method is effective.

Abstract:There are different methods combining misuse and anomaly detection for intrusion detection. However, most of them consist of more than one basic models which complicate the learning process. In this paper, we present an effective intrusion detection method with low complexity on the basis of the end-to-end memory network to classify the network behavior data by taking advantage of domain knowledge. A matching module and a blending module are designed in our model to ensure that relevant knowledge items take effect in the classify module. Furthermore, additional output are provided with the detecting result as explainable reference information. Data pre-processing is done using data normalization and knowledge items about attacks are selected from the dataset. Experimental results show that the domain knowledge plays a positive role in the model and the proposed method has good performance on intrusion detecting..

Abstract:In this study, Convolution Neural Network (CNN) is applied to video comprehension, and a driver fatigue detection algorithm based on multi-facial feature fusion is proposed. In the study, Multi-Task Cascaded Convolutional Neural Networks (MTCNN) is used to locate the driver's mouth and left eye. CNN is used to extract the static features from the driver's mouth and left-eye image, combined with the dynamic features that CNN extracted from the mouth and left eye optical flow to train for classification. The experimental results show that this algorithm with an accuracy rate of 87.4% is better than only use the static image for driver fatigue detection and it can well distinguish between yawning and speech actions that are similar in static images.

Abstract:The adaptive median filtering algorithm can effectively filter the impulse noise of image, however, with the noise density increasing, its filtering performance decreases progressively. For the improved median filtering algorithms of current, there are also relevant limitations. Against the limitations of the median filtering algorithm, an improved adaptive median filtering algorithm is proposed. It does noise detection based on the gray extremum of the filtering window. And it replaces the noise point with the gray median of the filtering window. If the gray median is noise point, it increases adaptively the filtering window to take a new gray median. If the filtering window has increased to the maximum size of allowed, and the gray median is still noise point, it takes the gray mean of the pixels except the gray extremum in the filtering window. Simulation experiment has been carried out for standard image and medical image, the results and datum of the filtering experiment demonstrate that, with the noise density increasing, the filtering performance of the standard adaptive median filtering algorithm decreases progressively; and the filtering performance of improved adaptive median filtering algorithm is still good, it maintains well the edges and details of image while filtering effectively the noise.

Abstract:The users are classified by different membership degrees with fuzzy C-means clustering. A more accurate clustering effect has been obtained and the problem of low recommendation accuracy caused by hard clustering is solved. Aiming at the privacy leakage problem of recommendation algorithm, the Laplace noise is introduced into the fuzzy C-means clustering process, and the differential privacy protection based fuzzy C-means clustering recommendation is implemented. The experimental results show that the proposed algorithm can effectively improve the security of the recommended system with the good quality of the recommendation.

Abstract:Unity engine-based manufacturing simulation system occupies large memory when loading, resulting that the system is not fluent at run time and cannot run smoothly. In response to this phenomenon, we design a resource dynamic scheduling algorithm from the perspective of model resource loading. Firstly, the scene is recursively divided into a plurality of leaf nodes for storing through a quad-tree algorithm. Then, combined with the resource dynamic scheduling algorithm, taking the camera position as the center, the resources of surrounding nodes are defaulted instance and destruction of the prefab to complete the memory management. Finally, according to Unity's memory management mechanism, the dynamic scheduling and memory optimization of scene resources are designed and implemented.

Abstract:To solve the problem of slow convergence speed before reaching the global optimum and low precision of optimization in Grey Wolf Optimizor (GWO), a hybrid GWO algorithm based on fuzzy weight strategy is proposed. By replacing the linear convergence factor in original algorithm with a new non-linear convergence factor, global search ability is improved. Furthermore, the algorithm employs a fuzzy weight strategy to offer discrepant weight to agents who are responsible for the decision, which will enhance the optimizing ability therefore. Numberical experiments are conducted in 23 standard test functions. Experimental results show that the proposed FWGWO algorithm has better performance compared with other algorithms.

Abstract:In view of the high time cost of traditional convolutional neural network, an improved convolutional neural network is designed, which has a reduction in the number of the convolutional kernels and an increase of the pooling methods. To solve the road traffic sign recognition problem of autopilot system and auxiliary driving system in the real scenario, the improved convolutional neural network is applied to road traffic sign recognition for the purpose of identifying traffic sign in a relatively short period of time. Taking the graphic data set GTRSB, real traffic sign image data as a sample, the real traffic sign is identified, and the overall recognition accuracy reaches 98.38%. Experimental results show that this method can reduce the recognition time while maintaining high recognition accuracy.

Abstract:For the dimensional disaster and feature redundancy problems in the process of data mining, a reinforcement learning based feature selection algorithm, which is combined Q learning methods with traditional feature selection methods, is proposed in this study. In the proposed method, the agent acquires a subset of characteristics autonomously through training and learning. Experimental results show that the proposed algorithm can effectively reduce the number of features and has higher classification performance.

Abstract:Personalized service is the inherent requirement and key point in building an intelligent learning environment. The utilized probability of learning resources can be improved by pushing algorithm for main body (learner) of the learning environment, and then can solve the problem that learners easily lose when they are studying on-line. The internal structure characteristics of the learners and learning resources are established through the unity semantics based on knowledge ontology, then a recommendation algorithm which combines the time attenuation function and difficulty matching method is designed to effectively calculate the correlation between them. The time attenuation function expresses the time-ordered behaviors of the learners in order to reflect the knowledge migration feature, and the difficulty matching method matches with learners' cognitive level and resource's difficulty. Finally, experimental results show that the time attenuation function and difficulty matching method reach the expected target and can guarantee the quality of personalized learning resources recommendation better in their common effect.

Abstract:In recent years, object detection is transferred to other fields, for example, face and vehicle detection. However, the bounding-box labeling is a huge resources cost work. This study solves the problem that transfer object detection task to other domain dataset without bounding-box label. A relationship layer is built to learn the relationship between classification and regression task. In addition, we construct a product dataset, on which rotatable object detection is solved using our training method. A proposal selecting method is proposed for training classification based on faster RCNN framework without bounding-box label. We propose a object detection method without bounding-box annotation. The method is easy to transfer to other datasets and training.

Abstract:Public Key Infrastructure (PKI) and SSL/TLS encryption are key elements of today's Internet for secure communications, but a major security risk is caused by an compromised or malicious CA. In 2013, Google proposed Certificate Transparency (CT) which aimed to safeguard the certificate issuance process by providing an open framework for monitoring and auditing HTTPS certificates. At present, in Google ecology, CT is being actively supported by most of CA, and developed in browsers. Meanwhile, a number of secure-related challenges remain. This article reviews the CT technology from the perspectives of trust mechanism and security threats, summarizes the CT-based Web-PKI trust model and security threat model, and puts forward the security assurance mechanism and application deployment recommendations. Finally, the development of CT technology is summarized and prospected.

Abstract:With the explosive growth of mobile applications, how to carry out UI automation testing efficiently and correctly becomes an important issue. Most of the traditional automated methods require developers to write test scripts manually, and another high-level testing method called "recording and playback" does not has capability of crossing devices. In addition, existing assertion mechanisms are generally not capable of describing the UI semantics completely. Due to those problems, this paper presents a new recording and playback method which is capable of crossing devices and describing the UI semantics. This method uses the widget path to precisely locate the widgets and employs the cross-device UI adaptive method to improve the capability of device-crossing. Furthermore, this study proposes two new assertion mechanisms to support UI semantics which check on number sorting and pictures. In addition, this study builds a prototype framework called RRF according to the proposed methods, and the experimental results show that RRF has a higher success rate of playback than other automated test tools.

Abstract:Remote sensing images play an important role in the development of species distribution model. However, low spatial resolution ecological niche derived from remote sensing data and lack of fine-scale presence data requires alternative approaches. With the application of various data acquisition devices, a large number of animal movement data can be considered as the presence data. In this study, we use DBSCAN method to cluster the movement data and each cluster represents a stopover. Then, we split the remote sensing image into 16×16 patches and divide them into positive and negative samples on the basis of clustering result. In addition, a multi-convolutional neural network model is proposed for the training and prediction of the potential distribution of wild geese surrounding Qinghai Lake. We evaluate the proposed system using a real GPS dataset collected on 29 birds over three years. The experiments show that the proposed method outperforms the GLCM method in terms of overall accuracy, F1 score, and AUC. The proposed method also can obtain a better result in a potential distribution prediction experiment.

Abstract:There are some associations between potential failures and running states of elevators. According to the key elements of elevators, the assessment model based on logistic regression algorithm is established by choosing assessment parameters and characterization reflecting the status of elevators. Through analyzing the evaluating model's basic principles, preprocessing original data, introducing the methods of penalty factor, cross validation, and high order pseudo linear, unbalanced data and difference issues are solved, and the evaluating model's accuracy is improved, thus the real-time supervision and pre-warning of elevators' status are established.

Abstract:SIFT algorithm is a classic method of image matching, but there are large amount of calculation and high time complexity. To solve these problems, we put forward an improved SIFT algorithm in this study. According to the eight gradient directions, we divided the 128 dimensional data of the SIFT algorithm into eight groups, and redefined the key point information. According to the new key point information, it generates new order descriptors. In this way, it will reduce the amount of calculated quantities, so as to improve the efficiency of the algorithm. The experiment shows that the improved algorithm keeps the advantages of the original algorithm, and greatly improves the efficiency of the algorithm without reducing the precision of the original algorithm.

Abstract:With the rapid development of Internet technology, the information data generated by all industries and professions is growing at an exponential rate. The traditional vehicle scheduling algorithm in dealing with dynamic vehicle scheduling problem, already cannot satisfy real-time and large-scale scenario, while big data in Hadoop technology can be a good solution. Therefore, this study constructs a dynamic vehicle scheduling parallel intelligent optimization algorithm based on Hadoop. Based on traditional genetic algorithm, the Hadoop platform parallel computing mechanism is used to improve the weak global optimization ability and converging to local optimal solution of the algorithm. The improved algorithm can effectively cope with massive and rapid response of the vehicle scheduling. The result of numerical calculation shows that the algorithm of vehicle scheduling based on Hadoop can effectively improve the optimization performance of traditional scheduling algorithm and has a good acceleration ratio when dealing with large-scale vehicle scheduling problems.

Abstract:Due to the internal details of the Oracle database is not open, it is necessary to resolve the Oracle network communication protocol TNS when performing security audits. The existing TNS protocol analysis level is not deep enough to cover the servers, clients, operating systems, and protocols versions. And common protocol reversing tools are not good at payload analysis. This study proposes a universal TNS protocol analytical solution according to common servers, clients, and protocol versions under windows and linux operating systems. Method of data mining is used for message segments with much bytes meaning unknown, and can gives fields that affect the message structure automatically. The application in the actual system shows that the proposed scheme can effectively analyze the large amount of data collected in the field, and extract the SQL statement from the request message.After the post-correction, all packets can be parsed without any exception.

Abstract:Netlink that implemented the interprocess communication function is a new protocol family in the Linux system. After referring to the Netlink mechanism of linux OS and implementation of 1553B protocol, and researching on the feasibility of combining sockets based on Netlink mechanism and 1553B protocol, this study designed the1553B protocol common socket based on Netlink mechanism. The socket API extends and defines the socket protocol stacks by Netlink common mechanism. It implements transmission of the1553B data using the socket API in Netlink mechanism. And it proves the convenience of the API through the experiment.

Abstract:In recent years, the Correlation Filter (CF) method in the field of tracking applications has made remarkable achievements. In this study, an effective occlusion detection mechanism and a scale transformation strategy are proposed to solve the problem that the relevant tracking is insensitive to the effect of target occlusion and the scale change. The tracking target is divided into four rectangular blocks with the center as the origin, and judging the degree of occlusion by calculating the Peak-to-Sidelobe Ratio (PSR) of the four blocks. And a new adaptive scale update strategy is proposed based on the position of the previous four peak response points. The method is tested on a public data set with problems such as occlusion, scale change, light change, etc. The simulation results show that the adaptive scale CF (OSCF) proposed in this study has good scale processing ability.

Abstract:Hackers and information disclosure have caused great network and information security threats to the campus network of colleges and universities. In the network security response mechanism and process, the rapid isolation and seizure of the accident point can greatly reduce the negative impact of the hidden risks on the entire campus network. Bearing this requirement in mind, this study has implemented a set of automatic management system by combining software and hardware. Using black-hole routing and OSPF routing distribution mechanism, and supporting script server and identity authentication system, we can isolate and seal external attack sources or internal failure point, thus promptly eliminate secondary hazards. At the same time, ACL control and OAuth authentication ensure the overall system security.