ISSN 0253-2778

CN 34-1054/N

2017 Vol. 47, No. 8

Display Method:
Original Paper
Design and implementation of a monitoring system for container-based cloudlet
ZHANG Song, SHU Guansheng, LI Jing
2017, 47(8): 627-634. doi: 10.3969/j.issn.0253-2778.2017.08.001
Cloudlet, which can perform resource-intensive applications for users with low delay, is a small data center located at the edge of the Internet. Compared to virtual machine technology, container technology has become the first choice to build a cloudlet due to its higher resource use rate, physical-machine-like performance and faster boot speed. The monitoring system is indispensable to the operation of a cloudlet. Combining the maturity framework of existing monitoring systems and the characteristics of container-based cloudlet, a monitoring system was designed and implemented for container-based cloudlets. And the monitoring system realizes the collection, storage, aggregating and displaying of monitoring data, anomaly detection, and dynamic settings of thresholds and collection cycles. At the same time, a request aggregating algorithm was proposed to reduce. Actual deployment and testing indicate that the system can meet the design requirements with low load.
A real-time multi-task scheduling framework in Linux user space
ZHANG Xu, GU Naijie, SU Junjie
2017, 47(8): 635-643. doi: 10.3969/j.issn.0253-2778.2017.08.002
Task scheduling in Linux kernel has tremendous overhead, and thus cannot fulfill the requirement of real-time applications. ULight, a real-time multi-task scheduling framework running in Linux user space was proposed to conquer this problem. ULight consists of three core modules: multi-task scheduling module, timer module and user-mode interrupt handling module. The multi-task scheduling module provides priority-based preemptive scheduling in Linux user space to reduce the overhead of task switch and scheduling; the timer module introduces a high-resolution timer system to support time-sharing scheduling, enable task sleep and increase preemption points; the user-mode interrupt handling module builds up an channel between kernel and user space, which enables user-mode threads to handle interrupts directly and efficiently in Linux user space. The experiment results show that, ULight brings much less overhead than Linux Pthread in terms of task scheduling; and keeps the precision of the timer stable within 20 μs; and can response to interrupts rapidly in user space.
An online outlier detection and confidence estimation algorithm based on Bayesian posterior ratio
SUN Shuanzhu, SONG Bei, LI Chunyan, WANG Hao
2017, 47(8): 644-652. doi: 10.3969/j.issn.0253-2778.2017.08.003
In order to satisfy the outlier detection requirements in one kind of high-speed, small-variance unlabeled industrial time series, an online outlier detection and confidence estimation algorithm based on Bayesian posterior ratio was proposed. The algorithm combined prediction and hypothesis testing, establishing the autoregressive model firstly and then using Bayesian posterior logarithm of residuals to identify outliers. To reduce misjudgment, the state transition probabilities were calculated by self-organizing map neural network and the reliability of detected outliers was evaluated afterwards. It updated models periodically to dynamically adapt to data changes, thus improving accuracy. Experimental results demonstrate that the online algorithm can effectively detect outliers in time series provide reliable confidence evaluation, bringing higher adaptability and practicability.
Features selection for video smoke detection using random forest
WEN Zebo, KANG Yu, CAO Yang, WEI Meng, SONG Weiguo
2017, 47(8): 653-664. doi: 10.3969/j.issn.0253-2778.2017.08.004
Using the random forest algorithm, a video smoke detection method with features selection was proposed. The method first extracted four original smoke image features including color features in RGB space, wavelet high frequency sub-images, multi-scale local max saturation, and multi-scale dark channel to input the random forest(RF). Then it utilized haze image formation model to make the synthetic smoke images from non-smoke images and partitions these images into blocks as the samples for RF. Thirdly, it trained RF to get the selected features from the original features and used support vector machine(SVM) to get a classifier which recognizes the smoke blocks and the non-smoke blocks. And then the smoke region candidate can be extracted from video images by the classifier. Finally, the method analyzed the detected smoke region with the features of the growth rate and the perimeter to area ratio to make the final decision on video smoke detection. The experimental results show that the proposed method can detect the smoke timely and give a fire alarm with a lower false-alarm rate.
Simultaneous localization and mapping based on RGB-D images with filter processing and pose optimization
XIONG Junlin, WANG Chan
2017, 47(8): 665-673. doi: 10.3969/j.issn.0253-2778.2017.08.005
RGB-D camera can capture color and depth images simultaneously, and is widely used for simultaneous localization and mapping (SLAM) research. In this article, The RGB-D SLAM method was improved from two aspects. Firstly, the point cloud filter method was improved to more effectively decrease the noise and redundancy of RGB-D camera data; secondly, an ICP algorithm was used to improve the estimated accuracy of the pose transformation matrix and the trajectories of camera movement. The proposed RGB-D SLAM method was verified on public datasets. The experimental results demonstrate that our RGB-D SLAM method can effectively improve the accuracy of the autonomous positioning and mapping of robots.
Magnetic resonance image reconstruction based on nonlocal augmented Lagrangian multiplier method
LI Chao, DU Hongwei, QIU Bensheng
2017, 47(8): 674-678. doi: 10.3969/j.issn.0253-2778.2017.08.006
Total variation (TV) is unable to recover the fine details and textures of magnetic resonance(MR) images since it often suffers from staircase artifact. To reduce these drawbacks, an improved TV MR image recovery algorithm is introduced by using nonlocal regularization into the CS optimization problem. The nonlocal regularization is built on nonlocal means (NLM) filtering and takes advantage of self-similarity in images, which helps to suppress the staircase effect and restore the fine details. On account of the complexity in implementing NLM filter, a modified MR imaging method called nonlocal Lagrange multiplier (MRNLM) is proposed to overcome the above shortcomings while boosting MR image quality. Experimental results demonstrate that the proposed algorithm shows significant improvements on the state-of-the-art TV based algorithms in both SNR and visual perception, as well as a fair balance between time and quality.
Joint antenna selection and cooperative communication design to enhance physical layer security
YAO Yanjun, ZHOU Wuyang, KOU Baohua, FAN Dandan
2017, 47(8): 679-685. doi: 10.3969/j.issn.0253-2778.2017.08.007
The prevailing research on physical layer security assumes that users are static, while takes no account of user mobility. In view of user mobility, a novel transmit antenna selection and cooperative communication design (TAS-Cop) to enhance physical layer security was proposed. The scenario considered contains two cooperative base stations, which are equipped with multiple antennas respectively. Firstly, both of the two transmitters apply TAS technology to maximize the signal to noise ratio (SNR) at the legal receiver. Next, when the legal receiver moves in the coverage of the two base stations, the transmitters cooperate with each other such that the best security performance can be acquired. Based on Rayleigh fading channel, key security metrics such as the closed-form expressions for non-zero secrecy capacity probability and secrecy outage probability were derived. Following that, an optimal power allocation strategy between the two transmitters was presented. Numerical simulation results show that the proposed scheme can achieve better security performance when compared with existing TAS schemes with user mobility is taken into consideration.
A two-stage feature selection method based on Fisher’s ratio and prediction risk for telecom customer churn prediction
XU Ziwei, WANG Peng, CHEN Zonghai
2017, 47(8): 686-694. doi: 10.3969/j.issn.0253-2778.2017.08.008
Telecom customer churn prediction is crucial to the customer relationship management systems of telecom operators. It aims to predict a particular customer who is at a high risk of churning. The predicting process includes the steps of data pre-processing, imbalance processing, feature selection, classifier training and evaluation. A two-stage feature selection method based on fisher’s ratio and prediction risk was proposed, which took advantage of the filter feature selection method and wrapper feature selection method to solve the high dimensionality problem of telecom customer churn prediction. The method was evaluated on a real-world dataset, and the experimental results verify that it is able to reduce feature dimensionality and improve the performance of classifiers.
Numerical solution to the Poisson equation under the spherical coordinate system with Bi-CGSTAB method
WEI Anhua, WU Qianqian, ZHU Zuojin
2017, 47(8): 695-698. doi: 10.3969/j.issn.0253-2778.2017.08.009
Solving the Poisson equations in spherical coordinate system is a key problem in computational fluid dynamics. Hence, using a conjugate gradient method named Bi-CGSTAB, the typical Poisson equation whose right-hand-side source term is-1 with zero boundary value was solved to give a numerical solution in a domain similar to that in round jet calculation Ω:{ r∈[7,52],  θ∈[-θb,θb], φ∈ [0,2π], θb=arctan(1/14)}. The numerical solution and its relevant residual of discretized equation were discussed.
Tool wear online monitoring of high-speed milling based on morphological component analysis
TAO Xin, ZHU Kunpeng, GAO Siyu
2017, 47(8): 699-707. doi: 10.3969/j.issn.0253-2778.2017.08.010
In high-speed milling, the cutter undergoes ultra-high-speed milling discontinuously, leading to rapid tool wear or breakage, which is difficult to monitor and will seriously affect machining accuracy and product quality, which underscores the importance of tool wear condition monitoring. Although the vibration method is an effective tool condition monitoring method, the vibration signal contains a variety of components and much noise, which decrease the accuracy of tool wear condition monitoring. To solve this problem, a sparse decomposition method of vibration signal was proposed based on the dual basis pursuit algorithm and morphological component analysis. First, morphological and sparse characteristics of the vibration signals in high speed milling were analyzed, and a dual basis pursuit framework was constructed and solved by an augmented Lagrangian variable splitting, thus separating the impulse components and harmonic components. Subsequently, two feature vectors, including the impulse density and amplitude
Lattice study in the pre-CDR of the SS ring for the injection chain of SPPC
XIE Kai, LIANG Linbo, REN Zhiliang, WANG Xiangqi
2017, 47(8): 708-712. doi: 10.3969/j.issn.0253-2778.2017.08.011
SPPC is of great significance to collider and the development of high-energy physics. As the last part of the injector chain, SS ring boosts the proton energy from 180 GeV to 2.1 TeV, and its performance is closely related to the beam quality of SPPC. The preliminary lattice study of the SS ring was focused on. Firstly, the filling pattern of SS ring was discussed based on the requirements of SPPC. Then, with MADX, two dedicated lattices, SSH and SSV, which were designed for fast longitudinal extraction scheme in horizontal and vertical directions were given in detail. The horizontal closed orbit deviations of SSH and SSV are smaller than 0.432 mm and 0.337 mm, respectively. The two lattices can provide a foundation for future research.