Computer and Modernization

Fine-grained Object Recognition Based on Part and Global Features

CHEN Shu-xian, LIU Jian-ming

2017, 0(10): 1-4,9. doi:10.3969/j.issn.1006-2475.2017.10.001

Asbtract ( 217 )

References | Related Articles | Metrics

Most fine-grained recognition only extracts global feature to classify and ignores visual differences of parts caused by attitude angle and pose. So this paper proposes a fine-grained recognition method by combining part feature with global feature. First, the paper does the target pose clustering to show the same visible part of object in the same pose, then extracts parts feature of object and combines global feature to classify in every pose. The proposed model is validated by the experimental results on the bird database CUB_200-2011, which has a significant effect on attitude and visual angle. The results show that the proposed method has better performance than the existing methods.

Apple Image Fusion Based on Scale-invariant Feature Transform

LUO Xiao-qing, WANG Peng-fei

2017, 0(10): 5-9. doi:10.3969/j.issn.1006-2475.2017.10.002

Asbtract ( 174 )

References | Related Articles | Metrics

A scale-invariant feature transform based image fusion method is proposed to detect the quality of apple in this paper. First, source images are decomposed into low frequency subbands and high frequency suabbands by nonsubsampled contourlet transform (NSCT). Second, a scale-invariant feature transform (SIFT) method is employed to find descriptors of low frequency subbands, which are used to construct a content matching metric. Next, this metric is introduced into the fusion of low frequency subbands. Then, a choose-max fusion rule is adopted to fuse the high frequency subbands. At last, the composite subbands are converted into a fused image by the inverse NSCT. Experimental results show the proposed method is sufficient and efficient in the application of apple quality inspection.

Segmentation of River Based on Self-supervised Learning

SUN Zhen, WANG Jing-dong, MAO Tian-yi, WEI Xue-ying

2017, 0(10): 10-14. doi:10.3969/j.issn.1006-2475.2017.10.003

Asbtract ( 170 )

References | Related Articles | Metrics

For the problems such as the complexity of the river because of the bridge images being caused by terrain, weather and environment, and hardly collecting all river samples of the images, a segmentation of river based on self-supervised learning is proposed. The approach uses the part of the river area automatically extracted by combining the K-means clustering method with Harris corner method as river sample in self-supervised learning, according to the color and texture feature extracted from river sample, trains the sample with the one class support vector machine. Then the river is segmented by the trained classifier. The experimental results demonstrate the proposed method has good performance in automatically segmenting river and can adapt to bridge images in different scenarios.

Image Segmentation with Detail Preserving Based on HMRF Using Neighborhood Selection

TIAN Qin-yi 1, TIAN Xiao-lin 2

2017, 0(10): 15-19. doi:10.3969/j.issn.1006-2475.2017.10.004

Asbtract ( 189 )

References | Related Articles | Metrics

Because of the filtering effect of the Markov random fields (MRF) region label model, detail structures may be preserved partially or lost entirely for synthetic aperture radar (SAR) image. The hidden MRF (HMRF) image segmentation approach with adaptive neighborhood systems based on scattering descriptor is proposed to better preserve detail features and border areas and to improve the segmentation effect. In order to improve the reliability and the adaptivity, fuzzy c-means (FCM) clustering algorithm is incorporated into scattering transform, where the choice of neighborhood shapes can be implemented adaptively. From among the different shape alternatives, the one with the highest fuzzy membership is chosen to compute the HMRF region label process. Experiment results demonstrate that the proposed algorithm improves the segmentation effect over the conventional HMRF using fixed shapes of neighborhood systems while detail structures are preserved well.

Recognition and Default Detection of Power Phase Sign Image

CHENG Jin-zhi, GUO Ling

2017, 0(10): 20-23,28. doi:10.3969/j.issn.1006-2475.2017.10.005

Asbtract ( 226 )

References | Related Articles | Metrics

Power phase signs are important electric facilities. Usually, they are installed in tower’s visible place. Withstanding rain and wind for long time, they are prone to damage, scratch, etc. Thus, a method of phase sign recognition and default detection is presented. Firstly, power phase sign image is processed by oriented filtering and graying. Secondly, Fixed threshold and FloodFill algorithm are used to segment the image. Thirdly, a matching method based on SURF operator and FLANN algorithm is adopted to recognize power phase sign. Last but not least, by calculating the percentage of intersection coefficient of B channel histogram, default can be judged. The effectiveness and robustness of the method are verified through experiments.

Data Fusion Technology Based on LEACH Protocol

HU Nai-ping, GENG Tong-tong, ZHOU Yan-ping

2017, 0(10): 24-28. doi:10.3969/j.issn.1006-2475.2017.10.006

Asbtract ( 242 )

References | Related Articles | Metrics

In this paper, a multi-level data fusion scheme based on LEACH protocol (MLDA-LEACH) is proposed. According to the hierarchy of LEACH protocol, first of all, the clusters use Kalman filtering algorithm to acquire the source data filtering to remove noise; secondly, the cluster nodes use distribution graph method to treat raw data to improve data accuracy, and then the adaptive weighted algorithm is used to aggregate the effective data. Simulation results show that MLDA-LEACH can effectively reduce energy consumption and prolong the life cycle of wireless sensor networks.

Research on Index Method of Massive Hydrology Data

FENG Jun, XU Wei-gang, FENG Du-qing, LU Jia-min, XU Huan

2017, 0(10): 29-35,41. doi:10.3969/j.issn.1006-2475.2017.10.007

Asbtract ( 160 )

References | Related Articles | Metrics

A large amount of hydrology data are stored in different forms and there are rich varieties of hydrology entity classes. For every type of hydrology entities, some basic description information and series of measuring business data involved in these entities are stored in different way with different update frequency. Hydrology business retrieve requests the index to provide basic descriptive information searching and a kind of combined query based on the relation between basic descriptive information and the business information. However, there is not an efficient index method which can consider several kinds of data and their dependencies. Furthermore, the rapid increasing of hydrology data also brings big challenges to retrieval performance. So, this paper proposes a distributed two-level index HRB based on Hadoop, which creates different index to satisfy different data types and retrieve requirements. The Experiments show that HRB is better at creating index than traditional distributed index, and when the amount of data reaches 10 million levels, HRB index retrieve data is faster. So, HRB has definitive value.

Design of Multi-source and Heterogeneous Data Fusion Framework for Integrated Lifecycle Management of EMU

ZHANG Chun, YUAN Tian-ning

2017, 0(10): 36-41. doi:10.3969/j.issn.1006-2475.2017.10.008

Asbtract ( 336 )

References | Related Articles | Metrics

With the continuous development of Chinese EMU industry, the EMU life cycle data management has become the focus of attention. Firstly, this paper analyzes the current status and existing problems of EMU’s life cycle data. Secondly, aiming at the characteristics of EMU’s multi-source and heterogeneous data, this paper proposes a data fusion framework based on EMU’s ontology knowledge base. Finally, this paper proposes a pattern matching method based on the similarity of composition patterns. Experiments show that the method has a good result of pattern matching of multi-source and heterogeneous data of EMU, and can greatly improve the automatization of multi-source and heterogeneous data fusion of EMU to a great extent.

Probabilistic Skyline Queries over Uncertain Moving Objects Based on Manhattan Distance

LI Jin-yang, CHEN Jia-liang

2017, 0(10): 42-48. doi:10.3969/j.issn.1006-2475.2017.10.009

Asbtract ( 201 )

References | Related Articles | Metrics

In many applications, due to the measurement instrument accuracy, update delay, network bandwidth and other restrictions, different forms of data uncertainty is widespread. At present, the information query in the uncertain data has been paid attention to by researchers in the field of database research, and it is also a hot topic to find efficient analysis method for uncertain data. This paper focuses on the uncertain moving object probability Skyline query problem based on Manhattan distance, and proposes a probability Skyline model based on Manhattan distance. Skyline model is used to solve the uncertainty of moving object at some point and obtain a p-t-Skyline result set. The result set contains all the moving objects whose Skyline probability is at least p at time t. In practical applications, the Skyline probability process for calculating a large number of uncertain moving objects is cumbersome and costly. In order to improve the computational efficiency of the probability Skyline query process, this paper presents a solution that includes four steps: sampling, bounding, pruning and refining. At the same time, a multidimensional index structure VCI tree is utilized to speed up the efficiency of data retrieval in order to further reduce Skyline computational overhead. The experimental results show that the solution has high efficiency on data sets of different scales and dimensions.

A Method for Determining Two Order Difference Cluster Number Based on Uniform Sampling

CHEN Yan 1, CHEN Guang 1, YI Ye-qing 2, LIU Qiang 1

2017, 0(10): 49-52,65. doi:10.3969/j.issn.1006-2475.2017.10.010

Asbtract ( 159 )

References | Related Articles | Metrics

Two order difference method for objective function uses it as a decision criteria that value of the objective function changes with the gradient of classes number. The two order difference algorithm directly uses the relation between the objective function value and the number of clusters to achieve the correct number of clusters on different data sets. But the calculation of the optimal cluster number will occupy a period of time. When the number of samples is large, the amount of calculation of using this method to obtain the optimum clustering number, will be also very large. To solve this problem, this paper proposes a method for determining two order difference cluster number based on uniform sampling. First, the improved uniform sampling design is adopted, and then the two order difference design is carried out on the subset of the data obtained. Experimental results show that this method not only can greatly reduce the amount of calculation, but also can achieve the desired correct judgment.

Speaker Verification of Normalization PLDA Based on T Matrix

GOU Xin-ke, WANG Yue

2017, 0(10): 53-56. doi:10.3969/j.issn.1006-2475.2017.10.011

Asbtract ( 172 )

References | Related Articles | Metrics

Recently, speaker verification based on i-vector/PLDA has become the state-of-the-art technique in speaker recognition.For the indefinite time speech, uncertainty of distortion and scaling, when i-vector with length normalization is converted to PLDA model, it affects the recognition rate. In this paper, the normalization of the length of the i-vector on the PLDA model is replaced by the normalization of total variability matrix T, to avoid the poor distortions. Experiments show that the method is similar to the length normalization, some of the results are better than that of the length normalization.

Fault Prediction of Wind Turbine Based on D-S Evidence Fusion

TIAN Yan-feng, LIU Shi-lei, JING Yan-jun, YANG Yi

2017, 0(10): 57-61,71. doi:10.3969/j.issn.1006-2475.2017.10.012

Asbtract ( 169 )

References | Related Articles | Metrics

Aiming at the mechanical and electrical faults of wind turbine generator, this paper presents a D-S fusion model based on electrical feature vector and vibrational feature vector. We construct two parameter-optimized support vector machines, as two evidences to predict the final fault pattern. Compared with the traditional fault diagnosis of generator for mechanical fault and electrical fault with vibration sensor and current sensor to distinguish different faults by spectrum characteristics, evidence fusion method can make current signal used for mechanical fault diagnosis, also vibration signal can be used for electric fault. Through a large number of experimental data analysis, fusion model compared with only a single signal structure has higher classification accuracy.

Pilot Design for NC-OFDM System Based on Cross Entropy Algorithm

XU Bi-ying 1, SUN Hui 2

2017, 0(10): 62-65. doi:10.3969/j.issn.1006-2475.2017.10.013

Asbtract ( 167 )

References | Related Articles | Metrics

Cross entropy is a kind of optimization method recently which shows good performance to solve combinational optimization problems. This paper, based on Minimum Mean Square Error (MSE) and cross entropy algorithm, proposes a kind of pilot design for non-contiguous orthogonal frequency division multiplexing (NC-OFDM) system. Firstly it generates samples of pilot position by Bernoulli distribution, calculates samples of MSE, and then updates distribution parameters by updating rule. Finally it gets reasonable pilot position. Simulation shows this method could achieve good MSE performance and bit error rate (BER) performance.

A Fast Fault Recognition Method of Brushless Synchronous Generator Rotating Rectifier

TANG Jun-xiang, CUI Jiang

2017, 0(10): 66-71. doi:10.3969/j.issn.1006-2475.2017.10.014

Asbtract ( 198 )

References | Related Articles | Metrics

Focusing on the slow speed problem of existing brushless synchronous generator rotating rectifier fault recognition methods, this paper presents a fast recognition technique based on improved extreme learning machine (ELM). The chicken swarm optimization (CSO) is used to optimize the parameters of ELM, and hence, an optimized model of ELM can be achieved, and then applied it to rotating rectifier faults recognition of brushless synchronous generator. Experimental results show that, the optimized ELM can achieve good diagnosis performance and high classification speed. The presented method can be considered to the application of brushless synchronous generator rotating rectifier faults recognition and localization.

Web Text Classification and Prediction Based on Thesaurus Match

YANG Yu-shi, HE Bo-xia, ZHOU Xin, LIU Hui-li, GE Fang-li

2017, 0(10): 72-75. doi:10.3969/j.issn.1006-2475.2017.10.015

Asbtract ( 224 )

References | Related Articles | Metrics

In order to achieve accurate classification of Chinese text, a classification method based on thesaurus match is put forward. The vector space model is used to express the features in the test set, the principle component analysis based on term frequency-inverse document frequency is used to weight the feature items in the corpus, 47 industries index thesaurus are screened out and built. And then the text industry category is determined according to the cosine similarity, the auto-regressive integrated moving average model is established, and the development trend of the next 10 days is forecast. Experimental results show that, the average classification performance F of index thesaurus is 85.6%, the average relative error of prediction model is 3.41%, which proves the classification method to be effective.

Micro-blog User Recommendation Based on User Profile

JIANG Zong-li， KANG Ya-ru

2017, 0(10): 76-80,86. doi:10.3969/j.issn.1006-2475.2017.10.016

Asbtract ( 239 )

References | Related Articles | Metrics

Using of single data source and the simple model in the traditional micro-blog recommendation results in the low recommendation accuracy. Therefore, a new recommendation algorithm based on labeled User Profile is proposed to overcome such issue. By analyzing the significance and correlation of individual user data, such as text, label, social relationship, and personal information, the algorithm generates new labels and suggests related interests by training LDA model and SVM classifier. The user’s interests are assigned by weighted sum of these factors. The overall recommendation accuracy is improved. The experiments show that the properties of the algorithm are better than the traditional VSM model, allowing users to have a better micro-blog experience.

Online Multiple Object Tracking Based on Parameter Learning and Motion Prediction

LI Peng-fei 1,2, LEI Hong 1

2017, 0(10): 81-86. doi:10.3969/j.issn.1006-2475.2017.10.017

Asbtract ( 170 )

References | Related Articles | Metrics

For short term occlusion and detector errors in online multiple object tracking, a new algorithm based on parameter learning and motion prediction is proposed. Firstly, the Kalman filter model is established by using the historical trajectory of the target, and target possible position in the current frame is given. Then, the cost matrix is established by calculating the correlation between the target and the current observation. The multi-target tracking is modeled as an assignment problem, and the Hungarain algorithm is used to solve the problem. In addition, the unusual situation of the target entering, disappearing and occlusion are processed. For the parameters of the multi-target tracking system, a new binary classification learning scheme is designed. Experimental results verify the effectiveness of parameter learning and the robustness against false detection, missed detection and occlusion. The proposed method has some advantages in many aspects compared with the performance of several classical algorithms.

Gaussian Sum-based Filtering for Hypersonic Object Tracking from Two Geosynchronous Satellites

CAO Ya-qin 1,2, QIN Ning-ning 1,2, YANG Le 1,2, LI Xi 3

2017, 0(10): 87-94. doi:10.3969/j.issn.1006-2475.2017.10.018

Asbtract ( 243 )

References | Related Articles | Metrics

This paper considers tracking a cruising hypersonic object with known altitude using the time difference of arrival (TDOA) and frequency difference of arrival (FDOA) measurements obtained at two geosynchronous satellites. A new tracking algorithm, referred to as GMM-AEKF, is proposed. The algorithm utilizes a discrete-time process equation under the WGS-84 ellipsoidal Earth model, which is established through discretizing the continuous-time equation of target motion with Euler sampling. GMM-AEKF generalizes the existing GMM-EKF algorithm in the sense that it implicitly takes into account the object moving along the earth surface with known altitude. It also includes a new method that can yield a more uniform GMM representation of the TDOA measurement, the alternative extended Kalman filter (AEKF) for FDOA track update, and a Kullback-Leibler (KL) divergence-based Gaussian component management scheme. Simulation results reveal that with the use of AEKF, TDOA and FDOA track updates of GMM-AEKF are the same as the state update of standard linear Kalman filter (KF). GMM-AEKF is also shown to be able to converge faster, which makes it more suitable for hypersonic object tracking other state-of-art benchmarks.

Image Encryption Algorithm Based on Life-like Cellular Automaton

PING Ping， HUANG Li-lin， MAO Ying-chi， XU Guo-yan

2017, 0(10): 95-99. doi:10.3969/j.issn.1006-2475.2017.10.019

Asbtract ( 174 )

References | Related Articles | Metrics

Concerning the problems of small key space, bad diffusion speed, low security of one-dimensional cellular automaton, as well as the extra rule storage space and unsatisfactory encryption effect of two-dimensional cellular automaton, we propose an image encryption algorithm based on life-like cellular automaton. Firstly, each pixel of the plain-image is converted into binary matrix. Then, the binary matrix is divided into two identical parts, which are taken as two initial states of the life-like cellular automata. Finally, a rule is selected as the key and the encryption is performed by the evolution of the life-like cellular automata. Experimental results show that the proposed image encryption scheme has a lot of characteristics, including large key space, and high sensitivity to the plaintext and key, which can effectively protect the security of the encrypted image.

Time-related Privacy Modeling and Consistency Checking

MA Wei-wei 1, JIANG Jia-xin 1, ZHANG Rui 2

2017, 0(10): 100-104. doi:10.3969/j.issn.1006-2475.2017.10.020

Asbtract ( 167 )

References | Related Articles | Metrics

With the rapid development of Internet, the issue of privacy protection is getting more and more attention. How to precisely describe the privacy requirement and guarantee the privacy requirement among different participants consistent with each other are two key issues in privacy protection. This paper proposes a declarative privacy requirement description language that supports time-related attributes. To verify the consistency of different privacy requirement, the mapping to the integrity constraint of the SCIFF frame is given. Finally, an example of online shopping is given to illustrate the feasibility of the method.

An Elliptic Curve Isomorphism Method for Resisting Differential Side-channel Analysis

WU Ke-ke, GAO Yue-fang, YAN Li-jun

2017, 0(10): 105-110. doi:10.3969/j.issn.1006-2475.2017.10.021

Asbtract ( 160 )

References | Related Articles | Metrics

Elliptic curve cryptosystems (ECC) are broadly applied in portable cryptographic devices. ECC provides the highest security strength per bit of any cryptosystem known today. However, such implementations of portable cryptographic devices of ECC are vulnerable to the widely known differential side-channel analysis (DSCA) attacks. Existing solutions reach the goal by increasing the computational costs, which prohibits the application of ECC in computation resource-restricted devices. Based on elliptic curve isomorphism mapping theory, an equal-value exchange model between elliptic curves is proposed, and then a security method that can prevent DSCA attack in ECC is designed, where almost does not increase computational costs of ECC. The accessment of security indicates that our method can resisit DSCA attacks.

The Third Party Management Platform Based on WeChat Public Platform

LI Peng-shu, FAN Hui

2017, 0(10): 111-115,120. doi:10.3969/j.issn.1006-2475.2017.10.022

Asbtract ( 289 )

References | Related Articles | Metrics

Considering the current application status of WeChat public platform and the universality of the use of WeChat public for the enterprises and individuals, making use of the functions and characteristics of the public platform, this paper makes the design and implementation process of the third party management platform on WeChat public platform. Through the analysis of the functional requirements of the third party management platform, this paper uses the Spring MVC, Spring Security, MyBatis, JSP and other Web development technologies to design a third party platform which manipulates WeChat public management. It is convenient for enterprise and individual users to control their public platform, effectively reduces the user workload and more concentrates on the business logic.

Nonlinear Energy Consumption Model of Android Application Based on Hardware Running Time

JIANG Hou-ming, HU Mu, CAO Hai-tao

2017, 0(10): 116-120. doi:10.3969/j.issn.1006-2475.2017.10.023

Asbtract ( 171 )

References | Related Articles | Metrics

The application of energy consumption analysis in Android application has always been an important part of mobile application testing. Based on the analysis of the hardware characteristics of mobile application and the underlying mobile terminal, a nonlinear energy consumption model based on hardware running time is proposed. Compared with the hardware energy consumption of complex use but high precision, the model lists the energy consumption of the single hardware in different states as the basic energy consumption unit, and then uses this time to describe the power consumption of the terminal in combination with the time variable, which can easily obtain and measure the running time, so as to quickly estimate the energy consumed by the application runtime. The results show that the average error of the energy consumption estimated by this model is less than 6% compared with the energy consumption measured by hardware measurement, which can provide an effective evaluation for the user to quickly detect the power consumed by the application.

Design and Implementation of USB 3.0 Communication Architecture Based on FPGA

WU Chun-chun, HU Huai-xiang, JIN Da, CHEN Xiang-yu

2017, 0(10): 121-126. doi:10.3969/j.issn.1006-2475.2017.10.024

Asbtract ( 315 )

References | Related Articles | Metrics

In order to solve the problem of bandwidth bottleneck in the communication between USB and host, a high speed communication architecture based on USB3.0 protocol is designed to provide an alternative scheme for USB data communication between embedded devices and PC. This design uses the EZ-USB FX3 chip of Cypress as a USB peripheral controller, FPGA as the main control chip in the hardware system. Through the design of FPGA hardware system, design and optimization of device firmware, the architecture supports USB 2.0/3.0 adaptive interfaces, it can realize the variable frame length communication between host and domestic embedded CPU and SRAM, hardware transmission speed of 360 MB/s, continuous data transmission speeds of up to 148 MB/s.

Table of Content