ISSN 0253-2778

CN 34-1054/N

2018 Vol. 48, No. 1

Display Method:
Original Paper
AR code recognition algorithm based on arc correction filter
YANG Meng, HE Yaxuan, YAN Deli, ZHANG Yong, WANG Weiming
2018, 48(1): 1-6. doi: 10.3969/j.issn.0253-2778.2018.01.001
AR codes are widely used in augmented reality (AR) systems and space vision positioning system, and their correct identification is a crucial technology to realize object identification and positioning in AR systems and space vision positioning systems. Classic AR code identification method performs best for AR codes on the plane, but poorly for AR codes pasted on cylindrical surfaces due to distortion. A method for recognizing AR codes is proposed. First, an improved AR coding method by adding location flags is used to encode the AR code, then the arc correction filter algorithm is used to correct the cylindrically distorted AR code images in two-dimensionally and three-dimensionally. The corrected AR code is then non-linearly partitioned according to the non- linear relationship between the rows of the AR code and the the length of each data bit. Finally the AR code identification is implemented by computing the pixel values. Experimental results show that the proposed method has a robust identification effect for cylindrically distorted AR codes.
Evolutionary algorithm portfolios based on information sharing
XU Han, LIU Weiming, LI Bin
2018, 48(1): 7-19. doi: 10.3969/j.issn.0253-2778.2018.01.002
A general framework for combining multiple evolutionary algorithms EAP_IS is proposed. Each of the constituent algorithms in this framework has its own population to maintain its characteristic and the continuity of the evolution process. EAP_IS runs each constituent algorithm with a part of the given time budget and encourages information sharing among the constituent algorithms. The effectiveness of EAP_IS has been verified by investigating 26 instantiations of it on 25 benchmark functions, and further comparisons of EAP_IS with other combinatorial frameworks have been conducted. Experimental results show that the proposed framework can improve the performance of constituent algorithms effectively.
A rule activation method for extended belief rule base based on improved similarity measures
LIN Yanqing, FU Yanggeng
2018, 48(1): 20-27. doi: 10.3969/j.issn.0253-2778.2018.01.003
When calculating negative individual matching degrees, there might appear negative values and all rules’ activation weights may be equal to zero. To address this problem, this paper introduces the Euclidean distance which is based on attribute weights and improves the traditional similarity computational formula. In addition, the traditional rule activation method activates all rules whose activation weights are greater than zero without considering inconsistency which exists in the activated rules, since the inconsistency of activated rules will weaken the reasoning performance of EBRB systems. Hence, considering the inconsistency existing in the activated rules, a new rule activation method of EBRB based on improved similarity measures is proposed. Compared with traditional rule activation method in the EBRB, the proposed approach activates rules by setting thresholds. And these activated rules are not only greater than zero but also have the smallest inconsistency. Finally, the pipeline leak detection problem and multiple public classification datasets have been employed to validate the efficiency of the new rule activation method. The experimental results show that the proposed method based on improved similarity measures can improve the reasoning accuracy of EBRB systems.
Locating failure-inducing combinations based on fault forest
WANG Yong, HUANG Zhiqiu, WEI Liangfen, LU Guifu
2018, 48(1): 28-34. doi: 10.3969/j.issn.0253-2778.2018.01.004
Combinatorial testing, a method for sampling parameter combination in the parameter space of a system, is suitable for systems in which failure is caused by a specific parameter combination. Based on the results of combination testing, locating the minimal failure causing schema (MFS) can help programmers to localize faults and repair them. However, combination testing might be affected by the mask effect, and even test cases containing MFSs may not necessarily trigger a failure. Therefore, it is extremely difficult to pinpoint MFSs in systems affected by the mask effect. A fault location method based on fault forest is proposed. Given a set of t-way combination test (t≥2) and their augment test set, this method first learns some basic fault trees which generate a fault forest, then extracts the basic suspicious MFS from the forest, and finally orders those suspicious MFSs by their suspiciousness which will help programmers perform further diagnosis. The simulation results show that the presented method can effectively identify MFS. In particular, for the systems affected by mask effect, result is robust.
Condition recognition of high-speed train based on multi-view weighted clustering ensemble
2018, 48(1): 35-41. doi: 10.3969/j.issn.0253-2778.2018.01.005
With the rapid development of China's high-speed train industry, some safety problems arising from the high-speed train operation are attracting more attention. Since the monitoring signals of the high-speed trains collected by sensors are nonlinear and non-stationary, it is difficult to identify the fault conditions of high-speed train. Therefore, in this paper, a multi-view clustering ensemble model based on weighted non-negative matrix factorization (WNMF) is proposed to it. Firstly, the vibration signals are analyzed the frequency domain, time-frequency domain and time domain. And the multi-views are obtained by extracting the eigenvector from the four aspects of the vibration signal, which are fast Fourier transform, wavelet packet energy, approximate entropy and fuzzy entropy of empirical mode decomposition, and the mechanical statistical characteristics. And then the clustering result of each view is obtained by the K-means. Secondly, two kinds of weight of the views are generated respectively by the contribution and the similarity of the clustering partitions. Finally, the output results of multiple clustering and the weights are combined for WNMF to ensemble. The experimental results show that the model can better identify fault conditions of high-speed trains.
Call admission control for multi-service heterogeneous networks
XU Ke, HUANG Hai, DONG Guangzhong, WANG Chuanqi
2018, 48(1): 42-46. doi: 10.3969/j.issn.0253-2778.2018.01.006
Call admission control, as an important part of resource management of heterogeneous network system, directly affects the effectiveness of the whole network resource usage. Therefore, it is very important to design a reasonable and effective admission control strategy in heterogeneous network environment. The call access problem in heterogeneous networks is considered. According to the bandwidth requirement of various services and the different profits they provide, a continuous time Markov analysis model for the random distribution of the call is constructed. A cache-based call access control algorithm is proposed and the iterative strategy is used to optimize the algorithm. The simulation results verify the effectiveness of the algorithm. The final control strategy is the optimal strategy for making the largest long-term system average profit.
A trajectory data density partition based distributed parallel clustering method
WANG Jiayu, ZHANG Zhenyu, CHU Zheng, WU Xiaohong
2018, 48(1): 47-56. doi: 10.3969/j.issn.0253-2778.2018.01.007
The development of global positioning technology and location-based service have contributed to the development of trajectory big data. Trajectory clustering is one of the most important trajectory analysis tasks and has been extensively studied. Currently, most of the clustering methods operate in a single-processor mode, and large-scale trajectory data processing is a lengthy process, making it difficult to meet the strong timeliness of the trajectory analysis task. To solve the problem, a distributed parallel clustering method based on trajectory density partition is proposed. Firstly, the whole dataset is abstracted in a rectangular region, and the dataset is divided into several partitions with tasks that have almost the same amount by the transformation of the longest dimension of the rectangle, thus constructing the local datasets for distributed parallel clustering. Then the worker servers implement the DBSCAN clustering algorithm for the local partitions respectively, and the manager server merges and integrates the local clustering results. The experimental results show that the algorithm is effective and improves the computational rate of clustering analysis to a certain degree.
A parallel algorithm for mining user frequent moving patterns
ZHU Yibo, BAO Peiming, JI Genling
2018, 48(1): 57-64. doi: 10.3969/j.issn.0253-2778.2018.01.008
Through daily moving trajectories, one can effectively find the frequent moving rules, i.e., user frequent moving patterns. Based on PrefixSpan algorithm, a parallel algorithm named PASFORM is presented for mining user frequent moving patterns. PASFORM uses a new pruning strategy to reduce the search space and several time constraints to make mining results time-tagged. It also employs the parallel method to mine mass data and a prefix tree to save the store space. Experimental results show that PASFORM is effective and efficient.
A calibrated lable ranking method based on naive Bayes
ZHANG Qilong, DENG Weibin, HU Feng, QU Yuan, HU Zongrong
2018, 48(1): 65-74. doi: 10.3969/j.issn.0253-2778.2018.01.009
The traditional calibrated label ranking algorithm (calibrated label ranking, CLR) uses pairs of label associations to transform and predict results. Its algorithmic calibration is achievely comparing it with the basis of binary relevance (BR). Its prediction has a certain dependence on the results of BR, thus incurring some limitations on the prediction of some datasets. To better distinguish between the relevance and irrelevance of the label, a method is presented for calibrating label boundary regions, which further corrects the boundary portion of the relevant label and the irrelevant label using Bayesian probability, thereby improving the accuracy of the classification of the boundary domain. CLR method based on naive Bayes(NBCLRM) presented is compared with seven traditional methods such as calibrated label ranking. Experimental results show that the proposed algorithm can not only adjust prediction results by modifying the thresholds ε and μ, but also effectively improve the performance of traditional multi-label learning methods.
A two dimensional semi-analytical model of sub-threshold surface potential analysis for fully depleted SOI MOSFET
CHANG Hong, SUN Guijin, YANG Fei, KE Daoming
2018, 48(1): 75-81. doi: 10.3969/j.issn.0253-2778.2018.01.010
Based on the principle of the SOI MOSFT, a definite solution of potential is proposed by introducing three rectangular sources in the oxide layer, depletion layer and buried oxide layer. The potential distribution of the three region has been obtained by means of the variables separation method, Fourier expansion method and the integral method. The solution is a special function of the infinite series expressions. The simulation results show that the proposed semi-analytical model has high precision and smaller calculation and can be applied to circuit simulation programs.
Position control of DC-motor based on sliding mode variable structure and high-gain observer
TANG Wenxiu, XI Wenlong, LI Zhipeng, WU Junying
2018, 48(1): 82-88. doi: 10.3969/j.issn.0253-2778.2018.01.011
A position sliding mode control method based on a high-gain observer is proposed for DC-motor position tracking control. The state space expression of the DC-motor is established. A high-gain observer is designed to observe the speed, current and their derivative signals. The accuracy and stability of the observer is guaranteed by placing suitable poles. The exponential reaching law of sliding mode variable structure is improved. The position tracking control of the DC-motor is realized by adopting the position