15.07.2014 Views

Conference Program of WCICA 2012

Conference Program of WCICA 2012

Conference Program of WCICA 2012

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Conference</strong> <strong>Program</strong> <strong>WCICA</strong> <strong>2012</strong><br />

discussed. And then, the wavelet transform is used to process echo<br />

signals, the low-frequency energy and the high-frequency energy are<br />

extracted as characteristics <strong>of</strong> echo signal in time-frequency domain.<br />

At last, a fuzzy pattern recognition algorithm is designed to evaluate the<br />

bonding quality <strong>of</strong> the composite material, it is based on the maximum<br />

membership degree principle. Simulation proved that the algorithm can<br />

recognize different echo signals quantitatively and effectively.<br />

◁ PSaC-03<br />

A Fast Stereo Matching Algorithm Used in Target Recognition <strong>of</strong> Mobile<br />

Robot, pp.4771–4774<br />

Yu, Naigong<br />

Lin, Jia<br />

Huang, Can<br />

Ruan, Xiaogang<br />

Beijing Univ. <strong>of</strong> Tech.<br />

Beijing Univ. <strong>of</strong> Tech.<br />

Beijing Univ. <strong>of</strong> Tech.<br />

Beijing Univ. <strong>of</strong> Techbnology<br />

For the accuracy and real-time requirements <strong>of</strong> stereo matching when<br />

mobile robot recognizes the target, an improved winner-take-all (WTA)<br />

algorithm which is based on the parallel binocular system is proposed.<br />

First, extract the relatively big difference points between two images<br />

as feature points. Then, implement the stereo matching for the feature<br />

points using WTA algorithm and only implement a simple verification<br />

for nonfeature points. Nonfeature points’disparity values are<br />

the ones <strong>of</strong> neighboring pixel. Finally, obtain a dense disparity map.<br />

The extracted feature points focus on the disparity discontinuity regions<br />

and the matching accuracy <strong>of</strong> the algorithm is equivalent to other existing<br />

algorithms.But calculation speed <strong>of</strong> the algorithm is faster and its<br />

edge feature is better.So it is a stereo matching algorithm with accurate<br />

matching and good real-time.<br />

◁ PSaC-04<br />

Smelting Process Smoke Detection using Multivariate Image Analysis,<br />

pp.4865–4868<br />

Zhang, Hongwei<br />

Song, Zhihuan<br />

Zhejiang Univ.<br />

Zhejiang Univ.<br />

To solve the problems that available technology was difficult to detect<br />

smoke concentration in the smelting process, this paper presented<br />

investigations into the smoke detection using multivariate image analysis.<br />

Firstly, smoke image matrix was transformed to a new one with<br />

single row and high dimension columns. Then the transformed matrix<br />

was processed by principal components analysis (PCA) to extract principal<br />

components (PCs). Finally, smoke concentration was detected by<br />

the two extracted PCs and threshold values. The correlations between<br />

PCs and smoke concentration were also identified and analyzed. The<br />

results <strong>of</strong> industrial application show that the proposed method could<br />

detect the smoke concentration effectively.<br />

◁ PSaC-05<br />

Optical Flow Estimation with Parameterized Data Term and Warping,<br />

pp.4633–4637<br />

Xu, Jintao<br />

Feng, Zuren<br />

Lu, Na<br />

Xi’an Jiaotong Univ.<br />

Xi’an Jiaotong Univ.<br />

Xi’an Jiaotong Univ.<br />

Horn/Schunck approach which has been widely used in variational optical<br />

flow estimation consists <strong>of</strong> a data term and a smoothness term.<br />

In this paper, a parameterized data term is proposed. The parameter<br />

can adjust the proportion <strong>of</strong> the two images’gradient for obtaining<br />

the Euler-Lagrange equations. This combination makes the optical flow<br />

more robust to noise and illumination changes. Firstly, the classic model<br />

is analyzed, especially the coefficients <strong>of</strong> the Euler-Lagrange equations.<br />

Then, a model with parameterized data term is proposed. This<br />

model has formulated the combination <strong>of</strong> image gradients theoretically<br />

and enabled more flexible combination. Finally, a multi-resolution technology<br />

has been used for solving the nonlinearity <strong>of</strong> optical flow. In this<br />

process, the warping step is also parameterized and the influence <strong>of</strong><br />

the parameter is analyzed. The experiments demonstrate the benefit <strong>of</strong><br />

our parameterized model to the classical one.<br />

◁ PSaC-06<br />

Hardware Design <strong>of</strong> The Wireless Automatic Meter Reading System<br />

Based on GPRS, pp.4536–4540<br />

Zhang, Ying<br />

Univ. <strong>of</strong> Anshan<br />

With the rapid development <strong>of</strong> computer network and electronic information<br />

technology, automatic and intelligent electronic products play<br />

roles that can not replaced by people themselves. Reading technology<br />

is a new technology that is applied in remote automatic data collection,<br />

transmission and processing for water, electricity, gas supply and management<br />

system, etc.. Hardware <strong>of</strong> wireless communication system<br />

based on GPRS was designed for remote automatic meter reading in<br />

this paper. The module supports UDP communication protocols, and<br />

can transmit remote data by the methods <strong>of</strong> messaging and network<br />

communications.<br />

◁ PSaC-07<br />

ANALYSIS OF MULTI-BIOMETRIC ENCRYPTION AT FEA-TURE-<br />

LEVEL FUSION, pp.4563–4567<br />

Fu, Bo<br />

Univ. <strong>of</strong> Electronic Sci. & Tech. <strong>of</strong> China<br />

The basic concepts and methods on performance and effectiveness<br />

evaluations at the feature-level fusion model <strong>of</strong> multi-biometric encryption<br />

are concentrated on in this paper. From the cryptographic theory<br />

point <strong>of</strong> view, firstly, the formal definitions related to mul-ti-biometric<br />

cryptosystems are formulated. Under some extreme conditions, the security<br />

and privacy <strong>of</strong> mul-ti-biometric cryptosystems at the feature level<br />

are ana-lyzed and rigorously proved. Finally, a close relationship between<br />

security and privacy and the fundamental trade<strong>of</strong>f between the<br />

accuracy and security are studied.<br />

◁ PSaC-08<br />

Classification Network <strong>of</strong> Gastric Cancer Construction based on Genetic<br />

Algorithms and Bayesian Network, pp.4676–4681<br />

He, Yiheng<br />

BeiJing Univ. <strong>of</strong> Tech.<br />

One <strong>of</strong> the most important link in improves diagnostic accuracy and<br />

disease cure rate is accurate classification <strong>of</strong> disease.The current gene<br />

chip’s development and widely applications making the diagnosis based<br />

on tumor gene expression pr<strong>of</strong>iling expectedto be on a fast and effective<br />

clinical diagnostic method. But the sample <strong>of</strong> gene is small and<br />

the expression data is multi-variable. In this article, we uses three data<br />

sets on gene expression pr<strong>of</strong>iles <strong>of</strong> gastric cancer for the construction<br />

<strong>of</strong> classification model, First, screened the gene which significantly<br />

changed in expression pattern, and use these genes as a set <strong>of</strong> the<br />

feature to reduce the number <strong>of</strong> variables, and then using genetic algorithms<br />

and bayesian network model to build the classifier, the build<br />

process uses these three gene expression data to learn classifier. Classification<br />

accuracy is calculated by leave-one cross-validation (LOOCV)<br />

and it reached 99.8%. Last we use the GO and pathway to analysis the<br />

classifier’s network structure.<br />

◁ PSaC-09<br />

Distributional Clustering Using Nonnegative Matrix Factorization,<br />

pp.4705–4711<br />

Zhu, Zhenfeng<br />

Ye, Yangdong<br />

Zhengzhou Univ.<br />

Zhengzhou Univ.<br />

In this paper, we propose an iterative distributional clustering algorithm<br />

based on non-negative matrix factorization (DCMF). When factorizing a<br />

data matrix A into CXM, an objective function is defined to impose the<br />

conditional distribution constraints on the base matrix C and the coefficient<br />

matrix M. It has been observed that, in many applications, the<br />

conditional distributions <strong>of</strong> instances are <strong>of</strong>ten employed to normalize<br />

the data dimensions. Taking these factors into account, we simplify the<br />

existent updating rules and obtain the iterative algorithm DCMF. This algorithm<br />

satisfies the constraints described above on condition that the<br />

instance matrix is preprocessed as a conditional distribution. DCMF<br />

is simple, effective, and only needs to initialize the coefficient matrix.<br />

As a result, the base matrix can be viewed as a centroid matrix and<br />

the coefficient matrix just records the membership <strong>of</strong> fuzzy clustering.<br />

Compared with several other factorization algorithms, the experimental<br />

results on text, gene, and image data demonstrate that DCMF achieves<br />

8.06% clustering accuracy improvement, 35.08% computational time<br />

192

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!