Computer Science and Artificial Intelligence

The following sections are included:

• Preface
• Editors
• Keynote Speaker
• Committees
• Contents

This paper realizes stereo matching of multiple views from arbitrary visual angles and gives a high-precision stereo matching algorithm under the premise that image correction is not made (the epipolar lines are not parallel), while a complete multi-view stereo matching algorithm from arbitrary visual angles is designed. Experiments have demonstrated that the algorithm in this paper has high precision and strong reflexivity, able to meet the smoothness constraint and obtain the overall optimal effect. The dense depth graph obtained by use of this algorithm can be used in three-dimensional model rebuilding.

Infrared gas sensor detecting gas concentration inaccurately is mainly because infrared gas sensor is susceptible to temperature, wind and other factors. This paper proposes an improved RBF neural network which can substantially overcome the environmental influences. Its idea is to optimize the parameters of RBF neural network by gradient descent method. Experimental results show that compared with RBF neural network the error of the improved RBF neural network is smaller.

Basic shape features of objects in a binary image are very important for image analysis and pattern recognition. In conventional algorithms, for basic shape feature computation, a labeled image generated by a connected-component labeling processing is usually necessary. This paper proposes a fast two-scan algorithm for calculating shape features of objects in a binary image without the use of a labeled image. Experiments demonstrated that the proposed algorithm is much more efficient than conventional algorithms for calculating basic shape features of objects in a binary image.

This paper, first briefly introduces the concept of network fire control system, communication network architecture oriented network fire control, and then a detailed analysis for the demand of communication network oriented network fire control for MAC. Then, according to these demands, proposes the innovation point of this MAC protocol design, and summarizes the design of MAC layer protocol. Finally, it analyzes the performance of the MAC protocol superior place and the need to further perfect it.

As a product of market economy, the stock market has its characteristics of high stakes and attractive benefits. But the fluctuation of stock price is always affected by many complexity factors. In this paper, China Merchants Bank stock is chosen as an example in utilizing good classification capability of BP neural network to forecast its future trend. In addition, the accuracy and stability of four kinds of neural networks in prediction of stock market trend are analyzed. Simulation experiments show that BP neural network has better performance than other approaches in learning capacity and generalization.

Up against the performance effect of the C4ISR system structure produced by cyber threats, a structure effect modeling method for C4ISR systems based on information flow was proposed. Firstly, the effect concept of C4ISR system structure was elaborated in this paper, and proposed lowering effect and damage two attacks and three attack strategies, and a structure information flow model was established. On this basis, the mathematical model of the damage degree for system structure based on the information flow and mathematical model of fallback effect for system structure based on the information flow delay is proposed.

Software defect prediction technology is designed to automatically detect whether the program module contains defects, so as to improve the quality of the software system. In view of the current LASSO feature selection method, limitations exist in the complexity metrics attributes selection, so LARS algorithm is added. This makes the complexity metrics attributes selected to be more concise and representative. The optimized metrics attributes are better in both the computation time and the software defect prediction capability.

The classical Dijkstra algorithm of graph theory only applies to solve single-source shortest path problem. Common Dijkstra algorithm will no longer apply if the shortest path must go through a specified set of nodes. This paper presents an interpolation algorithm based on improved Dijkstra algorithm which can solve the problem of obtaining the shortest path visiting a given set of nodes. This algorithm uses the Fibonacci heap to improve ordinary Dijkstra algorithm, and causes the given set of nodes to find the best insertion position sequentially. The nodes are then inserted using optimized Dijkstra algorithm. Finally, the shortest path approach can be generated through the specified set of nodes. Experimental results show that this algorithm can find the path which has smaller weight in a relatively low time complexity.

One of the important parameters to ensure the safe operation of boiler is the boiler drum water level in the thermal power plant, which is also a difficult one to measure due to the actual constrained situation. So it has practical significance to improve the accuracy and reliability of the boiler drum water level alarm. Through the study of the boiler drum water level alarm system, this paper proposes a new method based on D-S (Dempster-Shafer) evidence theory combined with fuzzy function and neural network in order to prevent a malfunction leak seized false detection by the single diagnostic method. Firstly, the boiler drum water level has been fuzzed. Then BP (Propagation Back) neural network and RBF (Basis Function Radial) neural network are respectively used to determine the alarm type. Finally, the diagnostic results of two neural networks are combined by the D-S evidence theory to determine the ultimate alarm type. Through the simulation experiment, the accuracy of the boiler drum water level alarm is higher than that of the current method and is also higher than that of a single neural network diagnostic method.

A new method of one-dimensional barcode detection based on pattern recognition is proposed, which is the rotation invariant LBP features combined with HOG features used in barcode detection. Firstly, a rotation invariant LBP feature image is generated by the rotation invariant LBP operator acting on the original barcode image, and it turns the feature image into a series of sub-blocks which have the same size, to produce a one-class classifier (barcode detection model) aiming at detecting barcode by SVDD training from some small image blocks of barcode samples. When letting the rotation invariant LBP feature sub-image be the input of the classifier, it can get an output on whether this sub-image is a part of barcode. Therefore, when the model acts on all sub-blocks of a picture on detecting, a rough position of barcode on this picture will be known. Finally, using HOG features of each original image’s sub-blocks and a relevant search algorithm to obtain a precise position about barcode. Experiment shows that the precision of this method compared with some current popular algorithm about barcode detection is higher.

Against such characteristics as uncertainty of controlled object in the main steam temperature control of thermal power plants, a novel PID control strategy with radial basis function (RBF) network tuning based on quantum-behaved particle swarm optimization (QPSO) algorithm is proposed. The QPSO algorithm is applied to optimize the initial parameters of RBF network, thus achieving dynamic control of the main steam temperature. The proposed controller has a self-learning ability, which can strengthen the system of uncertainties adaptability. Simulation results show that the control system performance is obviously better than the conventional cascade control.

In this paper, the model of the full trailer and the fractional order BP network are introduced. The error of the relative yaw rate of the tractor and the trailer as the training error of neural network is put forward and the steering control of the trailer is realized by neural network. In view of the control problem, the angle of trailer is compensated and greatly optimized the path tracking performance of the trailer on the basis of BP network. Finally, the driving path of the tractor and the trailer is completely consistent.

This paper aims to demonstrate the visualization strategies in excellent course resources design, development and management. Firstly, define the meaning of visualization according to the literature, summarize relevant data, information, knowledge of the excellent course resources and the application in the field by visualization. To theoretically discusses in design, development and management through visualization and content of the theoretical basis. Visualization should be throughout the whole process of the construction of the curriculum resources. The visualization of the high-quality goods curriculum resources not only improves the individual learning effect, but also guarantees the quality of the scale of the curriculum resources. Finally, the field of the practice of the fine course “computer culture basis” is based on the function of the visualization of high-quality goods curriculum resources optimization, and from the aspects of universality, collaborative and iterative visual strategy is proposed.

Space radiation contains various types of radiation, such as electrons, protons, gamma rays, and so on. In order to evaluate the adaptability of erbium (Er) doped fiber in the space radiation environment, a radiation source (Co⁶⁰ gamma ray source) is used to generate the dose rate and the total amount of radiation which is approximately equivalent to the ones via the space environment. By measuring the attenuation of the optical fiber for unit length in 1550 nm, the attenuation of the optical signal through the erbium doped fiber is measured. The radiation effect principle of the optical fiber is analyzed by the energy band theory and the energy level transition theory. The theoretical analysis and experimental results show that the attenuation of Er doped fiber is increased with the increasing of irradiation dose and tends to saturation.

Acousto-optic modulator (AOM) is one of the key technologies of coherent wind lidar system. By the theoretical analysis and the simulations, the mathematical model of the diffraction intensity and the diffraction efficiency of the AOM are obtained. The intensity of 0 - stage and 1 – stage diffraction beam change periodically with the increase of the ultrasonic power. When the wavelength of incident beam is 1550nm, the diffraction efficiency can reach about 84%, which is consistent with eye-safe.

With the popularization of accelerated test, how to get reliability and lifetime information accurately with limited samples has been more and more important. In this paper, a cyclic iterative algorithm of data and parameters to get more accurate parameters of accelerated degradation test (ADT) with a few samples is proposed. The method uses the parameters of acceleration model which is estimated by raw data to transform the data of stress at higher level to the data of stress at this level to increase the number of data in this level. Then, using the new data to estimate the new parameters until the parameters are not changed or changed negligibly. The maximum likelihood estimation (MLE) method is used to get more accurate parameters in each step of the cycle. Particle Swarm Optimization (PSO) algorithm is used to hunt for the optimum solution of the maximum likelihood function. In order to verify the validity of the method, a simulation model is built with MATLAB and an experiment of electric energy meter has been done. The proposed method is compared with the conventional methods in terms of their accuracy of parameters, which can fully demonstrate the effectiveness of the proposed method.

According to ship-borne early warning helicopters position deploy problem for the aircraft carrier in sea combat, a position deploy method is proposed. Based on analysis of the process aircraft carrier’s anti-sea combat, combined with combat background, the paper constructs the position deploy mode for early warning helicopter on the aircraft carrier’s anti-sea battle, including distance deploy model, bearing deploy model and height deploy model. Finally, using example to calculate the model simulation to verify it. The conclusion can provide useful reference for the aircraft carrier formation in the actual use of early warning helicopters.

Based on the experience and successes in the development of drilling information management software as well as the understanding of the South Azadagan well site situation in Iran, the South Azadagan Well Site Drilling Information Management System has been designed. Through the establishment of database of DIMS, the integrated well site drilling information management and application system is developed. In the process of collaborative information management under network environment, a convenient and unified data manage and analysis platform for well site drillers in different Rigs has been provided. This paper describes the design and development of South Azadagan Well Site Drilling Information Management System in detail.

Compared with relatively mature pedestrian detection, cyclist detection lacks extensive research for vulnerable road users (VRUs) protection. In this paper, Deformable Part Models (DPMs) for cyclist detection on Tsinghua-Daimler Cyclist Benchmark is proposed, including training strategies, data clustering, deformable parts and component elements. The key contributions that DPMs have in cyclist detection to further improve its performance is researched, and the optional configuration is 8 or 10 deformable parts with 3 or 4 flipped components. To solve its speed bottleneck, a fast cyclist detection method that speeds up ten times based on upper body proposal and cascade framework is proposed, which can be extended easily to other VRUs detection.

In order to deal with multi-dimensional functions optimization, an s-Particle Swarm Optimization(s-PSO) derived from the Local Siege Strategy was proposed. In s-PSO, each generation global optimal particle abandoned the traditional way to update the position instead of some siege particles was generated around it and then these particles approached it in a dynamic way to update its position. Experiments of several various multi-dimensional test functions were carried out to verify the validity of s-PSO and the results showed that they were able to handle with multi-dimensional function optimization with high success rate under 100 dimensionalities and its success rate was less than 50% for more than 100 dimensionalities.

This paper introduced the dynamic of electronic forensic identification both in China and abroad and proposed its categories and hierarchies. By analyzing the three legal bases and three technical foundations of electronic forensic identification, it indicated that this kind of examination could be both quantitative and qualitative, and its process could be precisely reproduced. This paper mainly discussed the legal scheme, technical scheme, and practical implementation plan in handling the case, as well as the identification working environment, and the supervision of the chain of evidence. The architecture of electronic forensic identification was proposed, which could unite the legal and technical aspects of this type of identification.

In order to evaluate and select the conceptual design solution better, the concept of trustworthiness is introduced, and proposes a conceptual design solution evaluation and selection algorithm based on trustworthiness. That is to get the comprehensive score by related trustworthiness computation, and then the preferable solution is ultimately selected. Finally, two experiments prove the necessity and feasibility of the algorithm, and it has a certain useful value.

In the transformer fault diagnosis, the span of the real amount of gas dissolved in transformer is greater than the training sample. Therefore, small sample training cannot meet the practical requirements of transformer fault diagnosis. But large sample size which is wide range and dispersion will reduce the generalization ability of neural network. Therefore, in view of the problems such as large amount of data and sample dispersion, it built the RBF neural network and GRNN (General regression neural network) neural network fault diagnosis model for the simulation results compared before and after the data standardization by using the cumulative frequency normalization processing on the data standardization and putting gathered all 441 groups of DGA data into network training and network detection. The results were as follows: cumulative frequency normalized data standardization method improved the neural network in different degrees of transformer fault diagnosis effect, and increased the value of practical application of transformer fault diagnosis.

In order to improve the efficiency and accuracy of transformer fault diagnosis, a new fault diagnosis method based on PCA and BP neural network is proposed. The proposed method used PCA to reduce the data of fault messages dimensions through matrix conversion, and then used BP neural network for transform fault diagnosis. Illustrated by examples, in the fault diagnosis of transformer, the PCA can concentrate fault information without influencing analysis results under the condition that the loss of useful messages is relatively small. Based on that, compared with RBF and GRNN neural network, the BP neural network has a higher accuracy for transform fault diagnosis. As a consequence, combination of PCA and BP neural network is a practical and feasible method in transform fault diagnosis.

Based on the distance measurement prototype, this paper proposes an angle error compensation method with monocular vision techniques for warehouse management system. To compensate the error of the distance estimation caused by pitch-down angle, the backward inference method for pitch-down angle calculation is given. Through analysis of factors affecting pitch-down angle calculation, a loop algorithm is designed for parameter optimization. The experiment result shows that the distance estimation error ratio is lower than 2%, which means that the optimal model is much better than the original one.

This paper proposes from the angle of data types and research contents in bibliometrics, Visualization as a widely used technology draws more and more attention in the field of bibliometrics. Some reasonable and effective visualization methods are proposed in accordance with the characteristics such as data types and the contents in bibliometric research through the way of analyzing Chinese surveying and mapping journals.

In realities, many project scheduling problems involve uncertain resources. In this paper, on the basis of the resource project scheduling problem, scenario sets are used to describe the upper limit of available resources and the uncertainty of arrival time and build a robust optimization model for URCPSP. Since the model involves uncertainties of double resources constrained scenarios and cannot be solved easily, a genetic algorithm is designed and an ant colony algorithm to approximate the robust solution. In the case analysis, the quality of two algorithms is compared which includes the optimal solution and the efficiency of solving, weighing the pros and cons, and solutions to the procedure.

The unknown vulnerability detection methods based on known vulnerabilities make an important significance. Since the code complexity, scientific extraction vulnerability related information is the key of the similarity matching methods to discover vulnerabilities. Dimensionality reduction based similarity matching function is one of the commonly used methods. However, faced with the choice of feature dimension reduction leads to bad arbitrary accuracy. To solve this problem, vulnerability discovery method based on the similarity of the program slicing, which slice through the promotion of a function expression vulnerability context information to enhance the accuracy of vulnerability discovery purposes. This paper also proposes a vulnerability detection framework based on program slicing similarity. It also solves the slicing feature points for vulnerability discovery labeling, automatic slicing etc. Thesis on open source projects Joern improves to achieve the above-described methods. Experimental verification shows that the method can significantly improve the accuracy of vulnerability discovery and improve the efficiency of vulnerability discovery, and demonstrates the effectiveness of the method and framework adopted.

Motion estimation is a crucial technology in dynamic scene analysis. Firstly, it estimates the motion parameters based on sub-block feature matching. Then, the motion parameters are utilized to compensate the global motion of the background. At last, it detects the moving targets by inter-frame difference in the stabilized image sequences. The algorithm is verified in different dataset. The results are analyzed qualitatively and quantitatively, which validate the improved performance of the proposed motion estimation algorithm.

For vehicle target, such information as motion direction of target, geometry main axis, and vehicle type are critical for target identification and classification. Local feature tracking can be conducted by using KLT-algorithm in the target region with the help of local feature, and vehicle target motion direction can be estimated by point motion. First extract strong response corner point in the target region based on images after compensation for global motion, then Local feature tracking can be conducted by using KLT-algorithm in the target region, and vehicle target motion direction can be estimated by point motion. The proposed method and various main axis has been compared in the experiment, and the result shows that the proposed method has the higher accuracy.

In order to estimate the system contribution rate of a certain equipment, the rough set theory and neural network are introduced into the evaluation of contribution rate of weapon equipment system, and it proposes the evaluation method of weapon system contribution rate combining with the rough set and neural network. It applies the training sample data set of rough set to simplify neural network, and eliminates the redundant data on the premise of keeping the important information. The simulation experiment shows that this method can achieve a better evaluation result. It takes the system contribution evaluation of a certain type of self-propelled artillery as an example. It constructs the evaluating model of the system contribution rate, the evaluation method of system contribution rate based on the rough set and neural network are presented and the basic steps and of the assessment method are also introduced.

The association rule mining is an important topic in the data mining. Association rule mining aims to find rules in the transaction database with the minimum support and minimum confidence which are user given. Apriori algorithm is one of the most influential algorithms for mining Boolean association rules, but it is ineffective because of large candidates and ineffective calculation for support. In order to find all the frequent item sets from the transaction database efficiently and quickly, an improved Apriori algorithm is presented to solve the bottleneck problems of the traditional Apriori algorithm. First of all, frequent item format is necessary, aiming to reduce the rate of memory occupancy in the process of frequent item generation, and then convert the storage structure of transaction records. Finally, it divides frequent 1-item into different groups in order to implement parallel computing. The experiment results show the improved Apriori algorithm can improve the efficiency of the original algorithm effectively.

In order to insure the validity of the escort fleet’s actions for collision avoidance at sea, it is important that considering every vessel’s requirement when adopt the optimal evasion course, and the fleet should keep original formation as far as possible. The paper introduces the basic principle of the fleet’s relative motion, then imports the conceptions of Formation Relative Movement Belt (FRMB), and uses the method to calculate the evasion course that refers to the datum vessel’s motion and the FRMB. A method is put forward to obtain the Distance of Closest Point of Approach (DCPA) and the Distance of Safe Point of Approach (DSPA), and a mathematical model is presented to decisionmaking of the optimal evasion course that fills the whole formation’s requirement. The results obtained lay a foundation for escort fleet’s collision avoidance.

In this paper, an effective hybrid optimization strategy is employed to simultaneously optimize the type of kernel function and the kernel parameter setting of Support Vector Regression (SVR), namely SVR-HPSOSA. This hybrid algorithm focuses on combining the advantages of PSO (fast calculation and easy mechanism) and SA (ability to jump away from local optimum solutions and converge to the global optimum solution). The HPSOSA algorithm combined the metropolis processes of SA into the movement mechanism and parallel processing of PSO. By combining the two methods, the PSO-SA algorithm has the advantage of both fast calculation and searching in the direction of the global optimum solution, helping PSO jump out of local optima, avoiding into the local optimal solution early and leading to a good solution quality. Compared with SVRHPSOSA and pure SVR, results show that the SVR-HPSOSA model can successfully identify the optimal type of kernel function and all the optimal values of the parameters of SVR with the lowest prediction error values in rainfall-runoff forecasting, can significantly improve the rainfall-runoff forecasting accuracy. Experimental results reveal that the predictions using the proposed approach are consistently better than those obtained using the other methods presented in this study in terms of the same measurements.

Measuring the period of a torsion pendulum is of great importance for most gravitational experiments. The improved fast Fourier transformation (FFT) method which was proposed by Barry G. Quinn can be used to determine the period of a torsion pendulum precisely. In this paper, several factors affecting the precision of the period estimation are discussed. The period of a sinusoidal signal subject to the window effect and the phase effect is estimated. Meanwhile, the effects are evaluated on the period determination due to other disturbances that would be present in a real oscillator, such as, higher harmonic, the damping, the monotonic drift and the white noise. The computer simulation experiments show that it is effective to overcome the disturbances mentioned above, and this method can be used to determine the period of a torsion pendulum with high accuracy.

In order to solve the problem of object tracking in practical application, this paper proposes a method of target tracking based on Bayesian framework. When the object appearance model is established, the background around the object is also used as a part of the model, which can solve the problem of appearance change and occlusion in a certain degree. In addition, since the phase of the image’s Fast Fourier Transform contains the texture information of the image, it is used to deal with the illumination change of object. Compared with the existing optical flow method, the usage of Fast Fourier Transform improves the real-time performance of the proposed object tracking method.

Aiming at the problems of Sybil attack group damaging routing protocol and positioning mechanism in ZigBee network, a Sybil attack group detection scheme is proposed. In the scheme, an available channel list structure algorithm and a Sybil group attack discovery algorithm was designed, and based on channel access fairness, Sybil attack groups was detected with anomaly traffic in the fixed time. Experimental results show that the scheme has a lower storage and computing overhead while efficiently detecting Sybil attack groups in ZigBee network.

DV-Hop localization algorithm is an important location algorithm for Wireless Sensor Networks. To solve the problem of large error on specific network, an improved algorithm about selecting shortest hop beacon node and computing the average hop distance is proposed. Simulation results show that it is useful for declining error rate of localization, and it is better for complex network especially.

As a classical problem in computer sciences, the multi-pattern matching algorithm has been widely researched and derives a lot of different matching algorithms. With the arrival of the age of big data and explosion of network information, original pattern-matching algorithm can no longer meet the need of practical application. This paper designed and implemented the parallelization of AC algorithm and WM algorithm based on GPU, proposed solutions to solve the problems of thread junction, optimized multi-pattern matching algorithm and tested the time consuming of these algorithm. The efficiency of coarse-grained thread-parallel AC algorithm is 20 times as much as serial algorithm, and improved when the size of test string and pattern set increases. The efficiency of coarse-grained parallel WM algorithm is 2 times as much as serial algorithm when the size of shortest pattern string is 2. At the same time, comparing the acceleration performance on GPU and FPGA, it further proves the superiority of GPU on pattern-matching algorithm.

With the research and development of software engineering, the industrialization production technology and method of software products have become the hot topics in software engineering field. The main problems facing software industrialization production are the research and implementation of software product line and PL-ISEE. In the paper, a new industrialized PL-ISEE model is proposed. One of the main parts framed the new PL-ISEE is the core asset and COTS component agent bus. To realize the agent bus and component based software assembly line, the broker idea and architecture based on CORBA are introduced, and the broker architecture and basic framework model adapted to the new PL-ISEE agent bus requirements are created, and then its CORBA based implementation mechanism and method are systemically discussed. The PL-ISEE realized by the broker architecture will have more advantages on the locating transparency, dynamic updates and expansion of the COTS servers, and interoperability and interactivity between different agent systems. These advantages are very useful to realize the new PL-ISEE and industrialization production of software products.

Brain extraction or skull stripping is an important pre-processing step in neuroimaging analysis. This review summarizes the current research efforts in the area of automatic or semi-automatic brain extraction from MRI volumes. The review firstly gives a full description of the various brain extraction methods based on categorization, then addresses the key problems (boundary leakage, local convergence and sensitivity to parameter setting) in brain extraction. The paper analyses how well these problems were dealt with in recent researches.

This paper proposes and designs a testing framework for the whole network based on the multi-layer software architecture. System based on TCP protocol and socket interface command flow, implementation of all kinds of network test equipment centralized control and status monitoring, improve the accuracy and efficiency of data transmission test system of anti-jamming in the field of communication.

Software reliability testing based on the traditional Markov chain usage model does not consider the software structural information, so it has a low ability of finding software defects. To solve this problem, a new reliability testing model is proposed in this paper. It contains the interaction information between software and environment, and the software structural information. The algorithm to generate test data set based on the Hierarchical Model is also provided. Testing based on the new model satisfies the software reliability testing request. It also improves the test data set’s ability of finding faults. Finally, a case study of demonstration software is presented to verify the effectiveness of the method.

Spectrum based fault localization techniques have shown promising results in assisting developers to find the possible locations of faults. These techniques employ the number of failed tests as well as the number of passed tests that cover statements to distinguish faulty statements and non-faulty statements. However, these techniques essentially assume the same importance for all the test cases, which ignore fault diagnosis ability for individual test cases. In this paper, an approach is proposed to quantify fault diagnosis ability for failed tests and passed tests. Two fault localization techniques are then proposed based on the approach to calculate the suspiciousness of statements for further ranking. Experimental results on Siemens Test Suite programs show that the proposed fault localization techniques are significantly more effective than traditional techniques.

Assessing the existing vulnerability in the target network, and establishing corresponding assessment model are the key to network vulnerability assessment. Aiming at the problem that the network vulnerability assessment involves many factors which are difficult to quantify, a kind of Hierarchical Network Vulnerability Assessment Model based on Attack Graph (AG-HNVAM) is proposed. This model combines with the Common Vulnerability Scoring System (CVSS), and adopts the idea of bottom-up and local-whole, which gives the vulnerability assessment value of three levels (service, host and network) intuitively. Experimental results show that this method is feasible and effective.

3D printer becomes popular and has great applications in many areas. An efficient placement algorithm in an available space of a 3D printer to print multiple 3D objects is important for the printing time cost. In this paper, a 3D model placement algorithm base on model’s contour is proposed. First, a robust algorithm is proposed to exacting the contour from 3D model. Second, the proposed algorithm adopts the bottom-left strategy combined with dynamic alter rotation to optimize the contour placement result. Moreover a tabu search is used to guide the search to get better results. Some 3D models are used to evaluate the algorithm, the experiment results show that the proposed algorithm can obtain optimal placement results.

This paper comes up with Rdp, a novel implementation of programming abstraction for R language under the circumstance of distributed parallel system. It keeps the traditional programming habits of R users and use MPI underneath to accelerate the execution performance. Furthermore, experiments are taken to prove the efficiency and usability of the new method.

The Satellite Condition Monitoring System (SCMS) plays a crucial role in ensuring the reliability and stability of satellite operation, however the design method of existing data processing software of SCMS requires the developers to redesign for a specified satellite model from scratch. Due to the different models and functional requirements between satellites, this development mode leads to a large amount of repetitive work, increases the development cost and declines the design efficiency. To overcome this development challenge, a component-based software architecture is proposed to achieve reconfigurable development of the data processing software. This paper first presents the unified component interface model, and then introduces the configuration description language to model the data processing flow. At last, a reconfiguration support software is designed to facilitate the developers to implement reconfiguration visually.

In order to improve the unsatisfied function of equipment attitude monitoring facilities, the equipment attitude monitoring system of the railway transportation based on WSN is designed. The framework of the software and the structure of the hardware had been studied and designed. The low pass filter has been applied to process the data of equipment’s attitude. The calculation algorithm of equipment attitude had been worked out and tested by using the models. Incidents caused by unexpected shift of equipment transported by train will be decreased, owing to this monitoring facilities.

With the development of the Internet of Things, the influx of mass sensed data, and the bottleneck of relational databases in functions, performance and stability, the industry’s leading IT companies have started to abandon the technical architecture of single relational databases, and the upgrade, transformation and migration of the software and hardware architecture of existing systems with a soaring size data for processing. Storage in the information systems of the transportation industry have also become a priority. Taking the upgrade and transformation project for the real-time probe vehicle system of Beijing for instance, this paper describes the real-time computing and storage management dilemma in the traditional architecture of the urban transportation core business system, new architecture solutions, and the effect of enhancement after upgrade and transformation.

A novel robust fuzzy controller design of nonlinear systems is investigated with assigning all poles into a specified disk, while the norms of input and output are bounded. Firstly, sufficient conditions are derived in terms of linear matrix inequalities (LMI) based on the Lyapunov theory. Secondly, a fuzzy controller based on parallel distributed compensation has been designed to meet all the desired control requirements via solving the corresponding sufficient conditions. Finally, control effect of a two-link robot system is proposed as a numerical example to illustrate the advantages of the proposed method.

In this paper, sliding mode controller with integral sliding surface is proposed to reject the system disturbances for permanent magnet synchronous motor (PMSM) speed regulation system. To reduce the chattering phenomenon, the signum function is replaced by the saturation function and fuzzy module is introduced in sliding mode controller. Simulation results show that the proposed control method can obtain satisfactory tracking performance and dynamic performance with smaller chattering.

The identification of charged particles with totally depleted Silicon Surface Barrier (SSB) detector under the deployment of front-side injection is more difficult than that of rearside injection, for the pulse shape of latter deployment is more sensitive to charge and mass of the detected ions than that of former deployment. Considering the successful application of the Artificial Neural Networks (ANNs) method on the discrimination of neutrons and gamma rays, the ANNs method is firstly introduced to identify charged particles with a totally depleted SSB detector under the deployment of front-side injection in the paper. Compared with the Pulse Shape Analysis (PSA) method, the experimental results show that the ANNs method is feasible and effective according to the value of Figure-Of-Merit (FOM).

With the rapid growth of the biomedical literature volume, the development of biomedical text mining technology is becoming more and more important. But the research progress of other semantic type such as disease has been hindered because of the lack of adequately availability annotated corpora. For all of disease centric information extraction task, the correct identification of disease entity is the key issue for further improvement. In this paper, a machine learning based approach that uses deep belief network as basic architecture, combining some simple orthographic features for disease mention recognition is proposed, which achieves comparable or better results than the state-of-the-art results on Arizona Disease Corpus. The paper also discusses why simple orthographic features can improve the performance under deep belief network architecture.

This paper presents a new structure-preserving method for color face recognition based on quaternion model. Inputting a new color face from the testing set, its projection is computed onto the chosen eigen-subspace, and then identify it by comparing its distance with the projections of known color face images. The processing is completed by a new fast structured method, in which only real computations are needed though the face images and the eigenface images are represented in quaternion forms. The experiments are conducted on color face databases: Faces95 and Georgia Tech face database, and the numerical results show that they achieved high level in both the performance and the efficiency in the proposed methods.

The AT89C51 microcontroller is regarded as the control center in intelligent alarm system for fire, and the system can receive and take a treatment on the concentration and temperature signal of smoke output by the fire detector with sound-light alarm. It can monitor the temperature and smoke concentration, etc. by sending inspection signal to the site continually, and have a feedback to alarm controller constantly. The controller compares the accepted signal with the normal value in storage to judge and determine whether there is a fire. When the smoke and temperature in site are anomalies, or fire occurs, it can realize sound-light alarm, the set of alarm limit of smoke concentration and temperature, self-diagnosis breakdown, delayed alarm, etc., which has a certain practical value.

This paper researches on the vibration characteristics of hydraulic excitation system based on the fluid-filled pipe and a hydraulic excitation testing system is established and an elastic plate connected with the pipe is served as a vibrating unit. Its mathematical model is established and the evolutionary process of the three-dimensional and two-dimensional vibration trajectories is simulated. By comparison, the vibration trajectories of the two-dimensional vibrating screen are regular, and that of the three-dimensional vibrating screen is confused, but which has higher screening efficiency. Serving as a new vibration mode, the three-dimensional vibration trajectory is based on the hydraulic excitation, has more advantages than that of the planar.

Face detection in complex background is widely used in face recognition, video retrieval, and human-computer interaction, etc. It is one of the hot topics in the study of computer vision and pattern recognition. A new method of face detection based on Relevance Vector Machine (RVM) and Support Vector Machine (SVM) is developed. First, the template matching based on average-face is used to eliminate a lot of simple background images, and then RVM classifier and nonlinear SVM classifier are combined to classify face and nonface images. The experiments show that this algorithm is effective.

Due to the traditional multi-exposure images fusion method performed in the RGB color space and cannot manipulate the intensity information conveniently, this paper proposes a new IHS color space based multi-exposure image fusion method, which transforms the input origin images from RGB color space to IHS color space and deal with the intensity and chrominance information separately. The fused intensity component is achieved by weighted sum in multi-resolution fashion and the weight map is estimated by the quality measures: local contrast and saturation. For the saturation and hue components, the best chosen pixel on the same location used the “winner-take-all” manner to recover the fused chrominance. Experimental results demonstrate the superiority and effectiveness of this method in term of subjective and objective evaluation.

Aiming at the positioning problem of a four legged robot, a mobile robot localization method based on robot binocular vision system and landmark information is proposed in this paper. The parallel binocular vision system is composed of two C270 cameras and a horizontal scale bar, and the mathematical model of the camera is established. The calibration experiment to the established camera model is done by using Zhang Zhengyou calibration method, to solve camera internal and external parameters. A target recognition algorithm based on color feature is proposed.

In this paper, a novel method is proposed based on a deep supervised autoencoder (DSAE) image reconstruction for face recognition. Unlike conventional deep autoencoder based face recognition method, the label information and the reconstruction residual information are taken into account in the learning procedure. First, an Adaptive Deep Supervised Network Template (ADSNT) with the supervised autoencoder is defined which is trained for labeled face images. Then, ADSNT is performed on face images to obtain the residual images which are computed by subtracting reconstructed images from original images. Furthermore, the residual images are applied by LDA to derive the compact discriminant feature representation. Finally, the features are used to train and test SVM for classification. The experiments demonstrate that the proposed method achieves promising performance.

Aiming at the high overhead cost in a single-chip system when the traditional internet security strategy is used in a rabbit farm environmental monitoring system. This paper proposes an improved scheme to realize asymmetrical bi-directional security authentication and key negotiation based on a third party in the system. The experiment proves that such a design scheme can meet the design requirement of bidirectional authentication and key negotiation between the IoT gateway and the client APP with the aid of the central server of the system without the need to increase the single-chip computing capacity of the gateway and the system construction cost.

The employment of mission load is very important to operational application of unmanned ship-based helicopter. Firstly, the actuality of unmanned ship-based helicopter and styles of anti-submarine load are analyzed, and then the fuzzy AHP is applied to resolve the problem. Furthermore, under the settled tactics background, the applied problems of anti-submarine load are optimized, and the feasibility and validity of the method are validated. It is a reference value to study on optimized choice of the mission load for unmanned ship-based helicopter.

An independent server program of drawing panel based on the original software system of SCARA2-DOF Googol robot is developed in this paper through utilizing the common module mode of Windows platform — DLL (Dynamic Linked Library). It is not only complex mathematic graph but also arbitrary hand drawing with mouse can be realized conveniently. This software operating system has good expansibility and widespread applicability. Besides complex mathematic 2-dimension graph, arbitrary hand drawing with mouse can be realized conveniently comparing with the original simple tutorial function.

In this paper, a big data novel filtering method is proposed– Local-loop Particle Filter Based on the Artificial Fish Algorithm (LPF-AF) for nonlinear dynamic systems. Particle filtering algorithm has been widely used in solving nonlinear/non Gaussian filtering problems. The proposal distribution is the key issue of the particle filtering, which will greatly influence the performance of algorithm. In the proposed LPF-AF, the local searching of AF is used to regenerate sample particles, which can make the proposal distribution closer to the poster distribution. There are mainly two steps in the proposed filter. In the first step of LPF-AF, extended Kalman filter was used as proposal distribution to generate particles, then means and variances of the proposal distribution can be calculated. In the second step, some of the particles move toward to the particle with the biggest weights. The proposed LPF-AF algorithm was compared with several other filtering algorithms and the experimental results show that means and variances of LPF-AF are lower than other filtering algorithms.

According to the feature of Buddha’s head light and backlight area is circle in Thangka image, so a method is proposed for detection of circle head light and backlight. Firstly, a method of morphology in edge extraction to get image edge is used, and by removing the edge connection point to get edge segment image. Secondly, through least square circle detection and verification condition of target circle to get the precise position of circle head light and backlight in Thangka image. The experimental results show that the method is with high accuracy and accurate positioning features in the detection of Buddha’s circle head light and backlight in Thangka Image.

In this paper, a pedestrian recognition algorithm of Multi-frame based on deep learning is proposed. First, a pedestrian image is adjusted to the same size, through image augmentation, make available information increases for different pedestrians. Second, a five layer CNN network is designed for pedestrian recognition, including two convolution layers, two pool layers, a whole connected layer, use softmax regression for multiple classes division in the last layer. Finally, experiments of different parameter settings are conducted, comparing with the Depth Belief Networks (DBN) and Depth Fully Connected Network (DFCN). The experimental results show that better pedestrian recognition rate can be obtained, even under insufficient pedestrian images or low resolution conditions.

The stock reviews are important professional advice for stock traders. Based on the massive stock reviews available online, an automatic approach which can identify the emotional tendency hidden in stock reviews will provide tremendous help to stock traders. A sentiment classification strategy is proposed for stock reviews analysis. Firstly, stock reviews are collected from the Internet and construct a labeled corpus. Secondly, a feature extraction approach is designed which can effectively map the corpus into a structured matrix. Finally, four classifiers are evaluated on multiple criteria. The synthetic oversampling approach is utilized to overcome the imbalanced distribution of positive and negative reviews. The experimental results demonstrate that the SVM algorithms coupled with SMOTE can achieve the highest performance for sentiment classification on the real world Chinese stock review corpus.

An image watermarking scheme is developed with echo-based hiding technique in the wavelet domain. In the embedding process, DWT coefficients with echo delay are utilized for watermarking. During the detection process, detect echo delay is first detected by the concept of image similarity and thus the embedded watermarks are extracted by the inverse procedure of embedding process. Experimental results confirm the watermarked image can keep high PSNR and high embedding capacity.

The research is designed to improve the result of Uyghur text classification. Based on current semantic text classification, this paper uses information gain and the KNN algorithm. Besides, it also provides an improved semantic text algorithm. And this algorithm has better results than the non-semantic algorithm.

Infrared imaging detection is widely applied in power equipment maintenance. In this paper, the temperature recognition based on infrared images of electric power equipment is proposed to obtain the temperature automatically to replace the time-consuming and error-prone manual way. Image enhancement and the morphological processing are adopted first to help the segmentation. Then, multi-segmentation based on prior knowledge is applied to get the temperature value region and the character region sequentially. The temperature value is obtained by the comparison of the character region and the standard character templates. The experiment is conducted on infrared images collected from the a power plant. The experimental results demonstrate the effectiveness and the practicability of the proposed method on automatic recognition of infrared image temperature of electric power equipment.

Intelligent UAV(Unmanned Aerial Vehicle) as a new field of artificial intelligence and robotics technology, which has widely received attention by the domestic and foreign industries. In view of the problem of the single function and the poor stability of flight that is widely known, this paper researches on the flight attitude of UAV based on IDE Arduino development platform, and the flight stability of UAV which is optimized through the PID control algorithm and fuzzy control method. At the same time, combined with the open source of Arduino, all kinds of sensors are installed on the UAV to carry out the underlying call, and finally, to realize the intelligent and stable flight control.

With the advent of the Chinese new BeiDou Navigation Satellite System (BDS), multisystem Real-Time Kinematic (RTK) high precision positioning service can be possible and available almost everywhere. This study aims to give an initial view of a BDS-based RTK positioning system combined with a path tracking algorithm. Application and tests will be carried out in Shanghai, China, on a real vehicle to analyze the feasibility of a high precision system to be equipped on an autonomous vehicle.

In order to monitor and alert the abnormal behavior of distraction, this paper proposes a method to detect the head behavior based on facial feature extraction. This method uses the ASM algorithm to obtain the facial feature points, calculates the head posture description on the position information of the face feature points, and draws the conclusion of head posture classified by SVM from the above information. Finally, experimental results show that the method proposed can effectively detect the abnormal behavior and the average detection rate is above 94%.

This study proposes an automatic fault location approach combining support vector machine. Different from the usual fault localization approach, the support vector machine is applied to classify the program statement into two classes. If there is only one class, the classification probability is used to rank the statements. If a statement has a minimum probability, it will have a maximum probability to be a fault. And empirical results of applying SVM are also presented to locate the fault and compare them against the results of other algorithms for the JTCAS program.

Finding the shortest loopless path of directed graph passing through specified nodes is the main purpose in this paper. The algorithm for the problem is a challenging work due to the fact that calculating the shortest path visits specified nodes is at least as difficult as the traveling salesman. To avoid duplicate search for invalid path in this paper, a new version of ant colony optimization metaheuristic is presented to ensure the extensive and convergence of the search, using priority queue to select paths and adopt Dijkstra algorithm to optimize the results of path which speeds up the generation of an optimal path. By utilizing ant colony optimization metaheuristic, priority queue and Dijkstra algorithm, the accuracy and efficiency of the algorithm is guaranteed, and the experimental results demonstrate the proposed method is effective.

In this paper, an omni-directional vehicle with mecanum wheel is designed for material delivery in the process of satellite assembly. Considering the character of mecanum wheel and environment of assembly hall, both machine vision navigation and magnetic tape are used on the omni-directional vehicle. An intelligent step length adjustment strategy is proposed for angle and displacement deviation, and is applied in the process of material delivery. Experiments demonstrate that the vehicle can deliver materials from warehouse to operation station automatically and accurately.

This paper aims at RFID problem, propose an Anti-Collision Algorithm based on Gray Code. This algorithm simplifies the reader to search tag prefix by Gray Code regulation and to enhance the identification speed through efficient division of the collision tags. Branch the tag colliding efficiently to enhance the identification speed of tag. The findings from analysis of algorithm and simulation result demonstrate that BSGC can reduce the number of collision and transmission delay, enhance throughput rate and increase the efficiency of tag identification.

This paper presents an investigation of reverberant Spoken Term Detection (STD) using frond-ends based methods. A 2-channel dereverberation method is adopted to achieve robust dereverberation under different reverberant conditions. A 2-channel spectral enhancement method is also used where the gain of each frequency bin is controlled by acoustic scene, which is detected based on the analysis of full-band coherent property. Deep Neural Network (DNN) as a feature extractor is used. The DNN based front-end allows a very flexible integration of meta-information. Bottleneck features are extracted in place of MFCC features used in HMM-GMM system which generates the lattice structure. The spoken terms are detected on lattice. The methods on the data provided by REVERB challenge are evaluated. On the corpus, the DNN front-end yields huge relative reduction in Equal Error Rate (EER).

An efficient color image encryption scheme with permutation-substitution architecture is proposed. The discrete generalized Arnold map is applied to disorder pixel positions and exchange the pixel gray values using exchanging strategy in the permutation process. The continuous generalized Arnold map is employed to generate pseudo-random sequences to realize the diffusion effect in the substitution process. One round of permutation and substitution can obtain perfect security effect. The security and performance of the proposed image encryption scheme have also been analyzed. Experimental results demonstrate that the proposed image encryption scheme is secure and effective for practical application.

To solve the problem of complex operations in the current smart home system, acquire a simple control method and increase feeling experience for users, the gesture recognition method based on Kinect was researched and integrated into the Human-Computer Interaction (HCI) system in smart home. Users can customize the gestures to realize the intelligent control of household equipment. The utilized template matching gesture recognition method is accomplished based on the Dynamic Time Warping (DTW) algorithm. The results of actual gesture recognition experiments show that the gesture recognition based on DTW is feasible and effective. The best identification distance is in the range of 2 - 2.5 m in front of Kinect and the highest recognition accuracy is reached on 96%.

According to the task planning problem of early warning system, the task planning model is set up, including the optimizing indexes of the tracking precision, task completing rate, resource laxity and sensor switching rate. A mixed Discrete Particle Swarm Optimization (DPSO) and Simulated Annealing (SA) algorithm is proposed to solve the task planning model. In the mixed algorithm, the ways of checking the restriction are proposed. Simulation results in the task planning instance show that the DPSO-SA algorithm is available and improves the planning quality.

Information extraction of visual attention can focus on the key areas of an image or video, which has important implications for target search, image search and other image processing algorithms. In this paper, a task and scale variable focus of attention based visual attention method is proposed. The proposed method top-down selects feature channels based on the key features of task target calculated by current task scenarios reasoning. Meanwhile, the scale variable round is used as focus of attention. The proposed method can have practical meaning, and universally meet the complex situations that the size of task target is not fixed in order to improve the hit rate of targets.

DNA sequencing technology has played an important role on life sciences, especially Illumina’s solexa sequencer. It was used for more and more transcriptome projects. In this paper, a tool named as RNASIM is developed, which can simulate solexa transcriptome sequencing based on the reference sequences. The tool RNASIM can actually simulate random gene expression and insert size of library with normal distribution in solexa pair-end sequencing. In order to validate the tool, three times of experiments were simulated based on the reference Arabidopsis transcripts and produced total reads’ length of 1Gb, 2Gb and 4Gb. Simulated reads were assembled and compared with original reference sequences. There were respectively 19,704, 20,376 and 20,692 transcripts aligned to one assembled contig. And there were only 1,071(3.2%), 849(2.5%) and 778(2.3%) original transcripts were not found related contigs in three assembled transcripts. Others were assembled multiple copy because of alternative splicing or gene duplication.

This paper proposed an automated repair method based on mutation testing, specifically for mutating code blocks of program. The code block mutation operator is designed and applied this mutation operator to repair defects caused by incorrect code formatting in C/C++ language programs. The empirical study was conducted and the results showed that 28,765 of 29,202 faulty programs caused by incorrect block of code formatting in the database were repaired automatically. The empirical result indicates that the designed mutation operator is highly effective.

Head detection plays an important role in pedestrian detection for the unique feature of human head. The existing methods of head detection mainly use the features of contour, color and template, which usually have low accuracy and robustness of detection. In this paper, a novel method based on CNN with multi-stage feature integration to detect head in complex background is proposed. This method combines the shallow and deep feature of CNN which can increase the completeness and comprehensiveness of image description. Meanwhile, the computation cost is decreased by reduced high dimensional features with PCA. The experimental results show that the proposed method has high detection accuracy, which is better than the existing methods.

Based on the Moving Finite Lines Source (MFLS) model, the heat transfer model for multi-boreholes fields was established, combined with quasi three-dimensional model of the circulating flow in the pipe of the borehole, the optimization model for multi-BHEs was established to guiding design. With the arrangement optimization, the utilization rate of the allowable area and the COP of the GSHP could be improved by 12.5% and 10%, respectively.

In this paper, a scheme for mitosis detection in breast cancer histological images is proposed. The whole scheme includes three main stages, images pre-processing stage, feature extraction stage and classification stage. In the candidate detection phase, morphological operation and marker-controlled watershed segmentation algorithm to generate candidates are performed. In the feature extraction stage, a total of 90 features including conventional morphological features and circular Gabor features of the candidate region is extracted and circular Gabor filter banks are designed for feature extraction. In classification stage, three different AdaBoost algorithms are employed to classify the candidates, the results shows that Modest AdaBoost achieves highest accuracy of 73.2%.

By using the studies of highway tunnel as research background, a tunnel lighting energy-saving LED lamp control system is designed, which according to the requirements of highway tunnel lighting safety, comfort level and energy-saving. The system uses monitor computer as the operation control core, and median machine as the local output controller. By using CAN bus control, going through the sensor module to collect vehicle flow rate, the speed and the brightness information. Subsequently, the system input control parameters as the control signal output to the lighting equipments from the operation control center are collected. Therefore, the command to the microcontroller and control the real-time automatic switch to on-off state in the different sections of the tunnel lighting is implemented. Based on the development of the lightings intelligence control system of CAN bus technology, to improve the lighting scalable class, the control precision, and the lighting continuity. In the meantime, it also saves the electric energy of the operation system. This paper includes the diagram for the tunnel control system, the research on the structure of the control system, and the usage analysis of the features and requirements of the control system.

Search engine, as a common tool in people’s daily life, can bring users better experience with its automatic error correction for the input, which is an indispensable function of the current search engines. This paper proposes an error classification-based method to correct the various query errors which usually happen in Chinese search engines by analyzing the online users’ query logs. To solve pinyin problem, training Hidden Markov model is explored. Experimental results demonstrate that appropriate methods in this paper acting on different error types can effectively improve the precision of error correction.

Base on the growing technology of LED lighting which combined with intelligent control method in tunnel lighting. The performance of long working life, low consumption and good light stability will definitely become the future trend. Through reasonable lighting, not only to ensure the driver can obtain the quality of visual information, but also guarantee the comfortableness and safety during driving in the tunnel. This paper mainly focuses on the research about mathematical calculation model of highway tunnel lighting system. Base on the studies of lighting computational scheme, in order to determine the correct calculation method. The result can be applied to guidance for designing the tunnel lighting system.

Financial management business includes modules of budget management, cost management, fund management, asset management, general ledger management and decision support. With the increase of Western Pipeline Company's business expansion, increasingly complex financial accounting and financial set of books appeared, doubling the workload and other issues. To solve these problems, it is proposed through the information technology to implement financial information management system of covering the entire business processes to further regulate the financial management and reduce human intervention. The system completes integration of automated processes in the greatest degree, improves decision support, data sharing and improving financial management.

Dynamic simulation model based on 1/2 of the vehicle is constructed by MATLAB software. With this model, centroid parameters of the vehicle body and elevation angle parameters are calculated. These data as inputs of the control systems, double fuzzy controllers is adopted in the active suspension system. Main power is controlled by elevation angle velocity and acceleration in front of the suspension system. Main force is controlled by centroid velocity and acceleration in rear suspension assembly. The results show that vehicle ride comfort and stability are improved significantly, through double fuzzy controllers, and optimizing these parameters of the body vertical acceleration, front and rear suspension stroke, elevation angle acceleration with performance rise by 23.8%, 47.0%, 67.3% and 23.8% respectively.

The competition intensity of internet access market has been ever-increasing in Shaanxi Province. Effective competition structure has not formed because of the backward supervision. Furthermore, internet access user loss was extremely serious because of many problems of upstream foundational network operators and downstream value-added service enterprise. In order to realize effective competition situation and promote vitality of the market, supervision department should relax restriction of market access and introduce competition, establish supervision system based on internet access service industry chain level, increase support for value-added service enterprise and strengthen supervision for interconnection, price and quality.

In this paper, the nonlinear SR system is introduced into the underground short-range sonic weak signal detection and successfully realized the drilling acoustic communication signal detection. It is better solved by the frequency modulation technology, introduced in the process of detection, that the SR system is not suitable for most impact attenuation signal and the SR system parameters are hard to match with the testing signals. Using stochastic resonance system has realized the drill pipe sound waves detection and parameter estimation under low SNR signal and the detection method breaks through the limitations of traditional detection method of weak signal by suppressing noise to improve the processing gain. Through the signal and noise and SR system to work together, it change part of the noise energy into signal energy, and produce resonance output similar to mechanics. This detection method greatly improves the effect of acoustic signal detection in weak signal environment and the purpose of the system outputting SNR to achieve identification of weak signals.

Horizontal well is an effective method to improve the recovery of tight sandstone gas reservoirs, but this kind of gas reservoir has strong heterogeneity with sandstone multiple stack and effective sandstone scattered development. During the implement of horizontal well, there are risk of lost circulation or sticking which will impair the gas field productivity. Besides, unreasonable design of well trajectories will cause insufficient development of reserves and loss of effective sandstone. In order to avoid the above phenomena, the geological study of tight sandstone reservoir need to be strengthened. Through long period field test of Sulige gas field, the differential geological design of unique well is determined. Then, the potential risk can be avoided based on well trajectories and the development of effective sandstone can be optimized. Using one well with the different horizontal well design ideas, from geological design to avoid risks. On the basis of abundant drilling geological parameters of Sulige gas field, this paper determines development characteristics of sandstone and effective sandstone using statistical method. Combined with previous research, three kinds of sand body stack types available for horizontal well development is proposed. Then on this basis, there are three kinds of well trajectories and effective sand body distribution characteristics. Base on this, three kinds of horizontal well trajectory and three-dimensional horizontal wells are designed to save the filed space and provide guarantee for improving the production efficiency of gas field.

This study proposed electronic Word Of Mouth (eWOM) conceptual framework for food products information adoption on the internet. Literature review reveals some streams that affect in eWOM involvement such as reviews, reviewers, websites, online opinion seeking, source credibility and their effects on the process of decision-making in the food industry. The results revealed that there was positive influence of reading some reviews because of product popularity, consistency, regency, usefulness rating of reviewers, and website’s reliability before information adoption on the internet. Likewise, eWOM involvement had a significant effect on online opinion seeking or leadership, source credibility and perceived risk. E-marketers need to encourage more on positive eWOM established discussion forums, emphasis on social networking sites, target the specific food products users to attract the customers, get consumers’ opinion, and increase the level of information satisfaction which can influence the purchase decision.

This article presents an illustration of moving track vehicles, builds an echo model of scatter points on the track, and analyzes impacts of various motion trends on radar echo. Moreover, it conducts simulation study in vehicles with different attitude angles based on micro-motion feature. Experimental results validate the effectiveness of this model.

With the development of software testing automation, the software test case generation based on modeling has become one of the key techniques. Now classic modeling has the limitations in the test case type and test case number. In this paper operation modeling is proposed to describe and model the use of the software in real life, based on which test case generation method is used to provide test cases of good quality for the embedded software.

Collaborative Filtering (CF) is one of the most popular algorithm used by many recommendation systems. However, there remain important research questions in overcoming the challenges such as cold startup, sparsity and poor prediction quality. In order to address these issues, Conditional Restricted Boltzmann Machines (CRBMs) which could take auxiliary information into account to collaborative filtering tasks are applied. The attributes of items are combined with the user’s behavior in the training of model and show that CRBMs could improve the accuracy of the prediction. Furthermore, based on the programming mode of CUDA, the parallel implementation is realized. The experiment results indicate that the GPU implementation can yield very good speedups (up to 47x) as compared with CPU implementation.

Normal modes method applies the technique of separation of variables to solve Helmholtz equations in computational acoustics, which is suitable for the sound fields with boundaries. The top boundary condition usually causes a singular case in atmospheric acoustics. The mirror boundary approach is that the sound field is placed, a mirror at the top to extend the original calculation area to an axisymmetric area with both rigid boundaries at the top and bottom. With a proper absorption coefficient assigned above the mirror position, the solution of the sound field below the mirror is correct. The approach is practical for normal modes method in atmospheric acoustics.

Pretty Fast Analysis (PFA) is an important tool for analyzing large-scale molecular dynamics simulation trajectory data. However, accurate analysis of large-scale molecular dynamics simulation requires a large number of simulation trajectory data and time consumption. Therefore, PFA is benchmarked for profiling and tuning PFA on heterogeneous systems composed of CPU and GPU. Benchmarking results demonstrate that it is advisable to start specified number of processes for ensuring correctness of running PFA on heterogeneous systems equipped with GPU. Furthermore, optimization including eliminating redundant synchronization and reorganizing special branch selection into computation would increase PFA performance by 7.1%.

To improve the sensitivity of statistical process monitoring while reducing costs, the economic design model of Autoregressive Moving Average (ARMA) process was built. Monte Carlo was used to calculate the Average Run Length (ARL) of ARMA control charts. Genetic algorithm was used to search for the optimal combination of parameters and the minimum cost of the model. Finally, the sensitivity analysis was carried out, which provides the parameters related to the expected cost in per unit time in the monitoring process in order to reduce the loss of quality.

Recently, along with logic theories being used in science research, various works have been advancing in auto-negotiation field, where multi-agent automated negotiation is a new popular research area. However, nearly all negotiation models used in research are based on the hypothesis that agent is rational and the importance of knowledge have not been taken into consideration. Negotiation process based on this model, which loses efficiency in solution, is complicated. This paper proposes a bounded rational negotiation model based on answer set program. It is to redefine and classify negotiation demand, propose the concept of core demand, and analyze the importance of literals in order to figure out the importance of demand as well as answer set dynamically. A negotiation method based on the concept of core demand is adopted in the model, where elements of the set of core demands must remain and only put non-core demands into negotiation. This method reduces the complexity of negotiation process and improves the solving efficiency.

To solve the problem of multi-source agricultural data integration and storage, the integration and storage models are imposed, and mix with different agricultural production data, for meeting a large number of users access to massive agricultural data synchronously. The method of agricultural data integration and storage with cloud services is proposed in this paper. Modern networks and cloud computing technology are used to resolve problems of varied information structures, information integration of different standards in different period of the experimental area. Experimental results of this method can prove the effectiveness of utilizing parallel file transfer method for agricultural data storage with cloud services.

Measuring the contribution of individual developers is very important for the governance of open source projects. In this paper, combining techniques of data mining and visualization analysis to design the contribution characteristics models of open source developers are attempted and study the differences between the models in software evolution processes that are hosted on GitHub. The experiment results show that the rhythm of the entire software evolution is driven by the activities of the small portion of core developers, who contribute to the project continuously rather than most contributors. Coding activities are also found to be closely accompanied by other communication activities such as issues and comments. These findings provide a solid foundation for contribution measurement of individuals in open source development.

This paper develops a new pedestrian detector based on margin theory. Support Vector Machine (SVM) with maximized minimum margin is widely applied in pedestrian detection. Recently, margin theory revealed that compared with the minimum margin, margin distribution is more critical to the generalization performance. In view of the imbalanced training data of pedestrian detection and important implication of the sample mean, this paper proposes Large Mean margin Distribution Machine (LMDM) which introduces the mean margin of two classes of sample means to remove the impact of imbalanced training data on the margin mean. LMDM maximizes the mean margin and minimizes the margin variance to regularize the model. Theoretical and experimental results show that LMDM can improve the pedestrian detection rate and have strong generalization performance.

The development of cloud services makes cloud software testing possible, and whether cloud testing in security has become a new focus. This paper combines the characteristics of cloud services and software testing, expounds the problems highlighted during cloud testing and analyses the risk based on the test. According to the security of the network, list the supplier and customer security lists, test encryption, authentication and authorization, formulate the corresponding specification and other aspects, proposes the strategy of the risk aversion in cloud testing and solve the security issues in the procedure of cloud testing to some extent.

Communication of voice over internet without encryption is no longer safe, a mobile application solution of secure voice over internet is proposed. The method is on the basis of combined secret key and hardware encryption which guarantees the security of voice communication for mobile phone users. The secure voice over internet protocol is accomplished with the support of smart card. Each secure voice communication over internet owns a different session key and encryption key. The session key is generated randomly, and the encryption key is generated on the basis of key seeds according to random numbers and time-stamp. It is hard to forecast and track the key generation, therefore illegal users cannot encrypt and decrypt the voice data for monitoring the voice communication.

In distributed manufacturing, manufacturing-as-a-service becomes a smart approach to realize cooperation between enterprises. Resource virtualization, which is considered as one of the key technologies, presents a flexible approach for users to use manufacturing resources in collaborative manufacturing systems. By analyzing the characteristics of manufacturing resources, a model of mapping from physical resource services to virtual resource services is proposed first. In the model, three kinds of mapping patterns are introduced to realize the virtualization. Then, the details of the components supporting the mapping are introduced. Finally, a case to show that the model is a practicable approach for resource service virtualization.

As smart phone is becoming more powerful, applications providing location-based service have been increasingly popular. Smart phone is commonly equipped with a variety of sensors, such as GPS, accelerometer, orientation sensor, etc. Although GPS can provide adequate location accuracy, it has limitations such as high energy-consumption. This paper presents an approach based on accelerometer, orientation sensor and GPS. It can reduce energy-consumption without compromising on location accuracy. Evaluation shows energy-saving of smart phone and location accuracy in a typical circumstance.

Driving behavior is reflected by the performance of the actual driver’s behavior in the driving process. It is an important basis for judging whether the driver is safe. Combining factor analysis, K-means algorithm, naive Bayes algorithm and k-fold cross-Validation algorithm, this paper builds a new driving behavior analysis model based on the movement data of the new energy bus. Then, it uses the real CAN and GPS data of the bus to validate the correctness and effectiveness of the model. The experimental results show that the model can classify the driving behavior excellently. It provides a good basis for standardizing driving behavior in the future.

The famous theory “Six Degrees of Separation” states that any two people around the world could connect with each other through an average number of six steps. In this paper, an attempt was made to analyze the Six Degrees of Separation in Baidu Encyclopedia which has included more than 1300, 0000 articles as of April 2016. A web crawler is built to get the links of articles in Baidu Encyclopedia and MySQL database is established to store the information in an architecturally sound way to make data analysis easy. The minimum number of links between any two unlikely articles was identified efficiently by using Breadth-First Search. According to the simulation result, even “Five Degrees of Separation” still holds in Baidu Encyclopedia; an article could get to 78% of all other articles within five links.

With the rapid development of Open Source Software (OSS), a lot of reusable software resources have been produced in open source communities. This leads to spotty quality and high dispersion of OSS resources so that lots of research works have been conducted for effective location of reliable software resources. To provide convenient data source for such studies, this paper introduces Octopus, a data acquisition system for resource of open source software. Octopus is a robust, scalable and efficient system which consists of three main modules that have been well decoupled and can be updated at runtime with flexible configurations. The experiment results show that, Octopus can successfully collect two kinds of OSS data sources: software production communities and software consumption communities. The former produce structured software artifacts such as project profiles while the latter contain rich user feedback such as posts and blogs.

The cloud-desktop based on Virtual Desktop Infrastructure (VDI) is deployed more often as the advanced mobile office solution. However, it is an assignable challenge to choose a fit VDI product at low cost for too many testing features to be investigated. In this paper, a multi-dimension factor decision-making model framework using back-propagation neural networks (MDMFBP) is proposed for the VDI-based cloud-desktop application evaluation. MDMFBP is highly data adaptive, applies and is able to account for correlation as well as interactions among features. This makes MDMFBP particularly appealing for high-dimensional cloud-desktop testing feature analysis. The experiment results show that our MDMFBP is workable, easy to implement and result in good estimation accuracies.

To increase the recognition rate of traditional method for license plate character, this paper introduces the Le-Net5 convolutional neural network of deep learning for license plate recognition. Firstly, this paper improves Le-Net5 and make it suitable for the particular task to continue network learning of character samples and optimize the parameters of the network. Finally, this paper tests the character of license plate using the trained network. Test showed that the recognition rate and anti-interference ability are greatly improved using CNN.

The following section is included:

• Author Index