Abstract: This paper presents a customized deformable model
for the segmentation of abdominal and thoracic aortic aneurysms in
CTA datasets. An important challenge in reliably detecting aortic
aneurysm is the need to overcome problems associated with intensity
inhomogeneities and image noise. Level sets are part of an important
class of methods that utilize partial differential equations (PDEs) and
have been extensively applied in image segmentation. A Gaussian
kernel function in the level set formulation, which extracts the local
intensity information, aids the suppression of noise in the extracted
regions of interest and then guides the motion of the evolving contour
for the detection of weak boundaries. The speed of curve evolution
has been significantly improved with a resulting decrease in
segmentation time compared with previous implementations of level
sets. The results indicate the method is more effective than other
approaches in coping with intensity inhomogeneities.
Abstract: It has proved that nonlinear diffusion and bilateral
filtering (BF) have a closed connection. Early effort and contribution
are to find a generalized representation to link them by using adaptive
filtering. In this paper a new further relationship between nonlinear
diffusion and bilateral filtering is explored which pays more attention
to numerical calculus. We give a fresh idea that bilateral filtering can
be accelerated by multigrid (MG) scheme which likes the nonlinear
diffusion, and show that a bilateral filtering process with large kernel
size can be approximated by a nonlinear diffusion process based on
full multigrid (FMG) scheme.
Abstract: Text categorization is the problem of classifying text
documents into a set of predefined classes. After a preprocessing
step, the documents are typically represented as large sparse vectors.
When training classifiers on large collections of documents, both the
time and memory restrictions can be quite prohibitive. This justifies
the application of feature selection methods to reduce the
dimensionality of the document-representation vector. In this paper,
we present three feature selection methods: Information Gain,
Support Vector Machine feature selection called (SVM_FS) and
Genetic Algorithm with SVM (called GA_SVM). We show that the
best results were obtained with GA_SVM method for a relatively
small dimension of the feature vector.
Abstract: Advances in processors architecture, such as multicore,
increase the size of complexity of parallel computer systems.
With multi-core architecture there are different parallel languages
that can be used to run parallel programs. One of these languages is
OpenMP which embedded in C/Cµ or FORTRAN. Because of this
new architecture and the complexity, it is very important to evaluate
the performance of OpenMP constructs, kernels, and application
program on multi-core systems. Performance is the activity of
collecting the information about the execution characteristics of a
program. Performance tools consists of at least three interfacing
software layers, including instrumentation, measurement, and
analysis. The instrumentation layer defines the measured
performance events. The measurement layer determines what
performance event is actually captured and how it is measured by the
tool. The analysis layer processes the performance data and
summarizes it into a form that can be displayed in performance tools.
In this paper, a number of OpenMP performance tools are surveyed,
explaining how each is used to collect, analyse, and display data
collection.
Abstract: Script identification is one of the challenging steps in the development of optical character recognition system for bilingual or multilingual documents. In this paper an attempt is made for identification of English numerals at word level from Punjabi documents by using Gabor features. The support vector machine (SVM) classifier with five fold cross validation is used to classify the word images. The results obtained are quite encouraging. Average accuracy with RBF kernel, Polynomial and Linear Kernel functions comes out to be greater than 99%.
Abstract: In this paper, we use Radial Basis Function Networks
(RBFN) for solving the problem of environmental interference
cancellation of speech signal. We show that the Second Order Thin-
Plate Spline (SOTPS) kernel cancels the interferences effectively.
For make comparison, we test our experiments on two conventional
most used RBFN kernels: the Gaussian and First order TPS (FOTPS)
basis functions. The speech signals used here were taken from the
OGI Multi-Language Telephone Speech Corpus database and were
corrupted with six type of environmental noise from NOISEX-92
database. Experimental results show that the SOTPS kernel can
considerably outperform the Gaussian and FOTPS functions on
speech interference cancellation problem.
Abstract: Discretization of spatial derivatives is an important
issue in meshfree methods especially when the derivative terms
contain non-linear coefficients. In this paper, various methods used
for discretization of second-order spatial derivatives are investigated
in the context of Smoothed Particle Hydrodynamics. Three popular
forms (i.e. "double summation", "second-order kernel derivation",
and "difference scheme") are studied using one-dimensional unsteady
heat conduction equation. To assess these schemes, transient response
to a step function initial condition is considered. Due to parabolic
nature of the heat equation, one can expect smooth and monotone
solutions. It is shown, however in this paper, that regardless of
the type of kernel function used and the size of smoothing radius,
the double summation discretization form leads to non-physical
oscillations which persist in the solution. Also, results show that when
a second-order kernel derivative is used, a high-order kernel function
shall be employed in such a way that the distance of inflection
point from origin in the kernel function be less than the nearest
particle distance. Otherwise, solutions may exhibit oscillations near
discontinuities unlike the "difference scheme" which unconditionally
produces monotone results.
Abstract: This project aims to investigate the potential of
torrefaction to improve the properties of Malaysian palm kernel shell
(PKS) as a solid fuel. A study towards torrefaction of PKS was
performed under various temperature and residence time of 240, 260,
and 280oC and 30, 60, and 90 minutes respectively. The torrefied
PKS was characterized in terms of the mass yield, energy yield,
elemental composition analysis, calorific value analysis, moisture and
volatile matter contents, and ash and fixed carbon contents. The mass
and energy yield changes in the torrefied PKS were observed to
prove that the temperature has more effect compare to residence time
in the torrefaction process. The C content of PKS increases while H
and O contents decrease after torrefaction, which resulted in higher
heating value between 5 to 16%. Meanwhile, torrefaction caused the
ash and fixed carbon content of PKS to increase, and the moisture
and volatile matter to decrease.
Abstract: In this paper we are interested in classification problems
with a performance constraint on error probability. In such
problems if the constraint cannot be satisfied, then a rejection option
is introduced. For binary labelled classification, a number of SVM
based methods with rejection option have been proposed over the
past few years. All of these methods use two thresholds on the SVM
output. However, in previous works, we have shown on synthetic data
that using thresholds on the output of the optimal SVM may lead to
poor results for classification tasks with performance constraint. In
this paper a new method for supervised classification with rejection
option is proposed. It consists in two different classifiers jointly
optimized to minimize the rejection probability subject to a given
constraint on error rate. This method uses a new kernel based linear
learning machine that we have recently presented. This learning
machine is characterized by its simplicity and high training speed
which makes the simultaneous optimization of the two classifiers
computationally reasonable. The proposed classification method with
rejection option is compared to a SVM based rejection method
proposed in recent literature. Experiments show the superiority of
the proposed method.
Abstract: An automated wood recognition system is designed to
classify tropical wood species.The wood features are extracted based
on two feature extractors: Basic Grey Level Aura Matrix (BGLAM)
technique and statistical properties of pores distribution (SPPD)
technique. Due to the nonlinearity of the tropical wood species
separation boundaries, a pre classification stage is proposed which
consists ofKmeans clusteringand kernel discriminant analysis (KDA).
Finally, Linear Discriminant Analysis (LDA) classifier and KNearest
Neighbour (KNN) are implemented for comparison purposes.
The study involves comparison of the system with and without pre
classification using KNN classifier and LDA classifier.The results
show that the inclusion of the pre classification stage has improved
the accuracy of both the LDA and KNN classifiers by more than
12%.
Abstract: A simple and easy algorithm is presented for a fast calculation of kernel functions which required in fluid simulations using the Smoothed Particle Hydrodynamic (SPH) method. Present proposed algorithm improves the Linked-list algorithm and adopts the Pair-Wise Interaction technique, which are widely used for evaluating kernel functions in fluid simulations using the SPH method. The algorithm is easy to be implemented without any complexities in programming. Some benchmark examples are used to show the simulation time saved by using the proposed algorithm. Parametric studies on the number of divisions for sub-domains, smoothing length and total amount of particles are conducted to show the effectiveness of the present technique. A compact formulation is proposed for practical usage.
Abstract: This work deals with aspects of support vector learning for large-scale data mining tasks. Based on a decomposition algorithm that can be run in serial and parallel mode we introduce a data transformation that allows for the usage of an expensive generalized kernel without additional costs. In order to speed up the decomposition algorithm we analyze the problem of working set selection for large data sets and analyze the influence of the working set sizes onto the scalability of the parallel decomposition scheme. Our modifications and settings lead to improvement of support vector learning performance and thus allow using extensive parameter search methods to optimize classification accuracy.
Abstract: In this paper we propose a method for recognition of
adult video based on support vector machine (SVM). Different kernel
features are proposed to classify adult videos. SVM has an advantage
that it is insensitive to the relative number of training example in
positive (adult video) and negative (non adult video) classes. This
advantage is illustrated by comparing performance between different
SVM kernels for the identification of adult video.
Abstract: Study of fire and explosion is very important mainly
in oil and gas industries due to several accidents which have been
reported in the past and present. In this work, we have investigated
the flammability of bio oil vapour mixtures. This mixture may
contribute to fire during the storage and transportation process. Bio
oil sample derived from Palm Kernell shell was analysed using Gas
Chromatography Mass Spectrometry (GC-MS) to examine the
composition of the sample. Mole fractions of 12 selected
components in the liquid phase were obtained from the GC-FID data
and used to calculate mole fractions of components in the gas phase
via modified Raoult-s law. Lower Flammability Limits (LFLs) and
Upper Flammability Limits (UFLs) for individual components were
obtained from published literature. However, stoichiometric
concentration method was used to calculate the flammability limits
of some components which their flammability limit values are not
available in the literature. The LFL and UFL values for the mixture
were calculated using the Le Chatelier equation. The LFLmix and
UFLmix values were used to construct a flammability diagram and
subsequently used to determine the flammability of the mixture. The
findings of this study can be used to propose suitable inherently
safer method to prevent the flammable mixture from occurring and
to minimizing the loss of properties, business, and life due to fire
accidents in bio oil productions.
Abstract: Linearization of graph embedding has been emerged
as an effective dimensionality reduction technique in pattern
recognition. However, it may not be optimal for nonlinearly
distributed real world data, such as face, due to its linear nature. So, a
kernelization of graph embedding is proposed as a dimensionality
reduction technique in face recognition. In order to further boost the
recognition capability of the proposed technique, the Fisher-s
criterion is opted in the objective function for better data
discrimination. The proposed technique is able to characterize the
underlying intra-class structure as well as the inter-class separability.
Experimental results on FRGC database validate the effectiveness of
the proposed technique as a feature descriptor.
Abstract: In this paper, we present an innovative scheme of
blindly extracting message bits from an image distorted by an attack.
Support Vector Machine (SVM) is used to nonlinearly classify the
bits of the embedded message. Traditionally, a hard decoder is used
with the assumption that the underlying modeling of the Discrete
Cosine Transform (DCT) coefficients does not appreciably change.
In case of an attack, the distribution of the image coefficients is
heavily altered. The distribution of the sufficient statistics at the
receiving end corresponding to the antipodal signals overlap and a
simple hard decoder fails to classify them properly. We are
considering message retrieval of antipodal signal as a binary
classification problem. Machine learning techniques like SVM is
used to retrieve the message, when certain specific class of attacks is
most probable. In order to validate SVM based decoding scheme, we
have taken Gaussian noise as a test case. We generate a data set using
125 images and 25 different keys. Polynomial kernel of SVM has
achieved 100 percent accuracy on test data.
Abstract: Intrusion detection is a mechanism used to protect a
system and analyse and predict the behaviours of system users. An
ideal intrusion detection system is hard to achieve due to
nonlinearity, and irrelevant or redundant features. This study
introduces a new anomaly-based intrusion detection model. The
suggested model is based on particle swarm optimisation and
nonlinear, multi-class and multi-kernel support vector machines.
Particle swarm optimisation is used for feature selection by applying
a new formula to update the position and the velocity of a particle;
the support vector machine is used as a classifier. The proposed
model is tested and compared with the other methods using the KDD
CUP 1999 dataset. The results indicate that this new method achieves
better accuracy rates than previous methods.
Abstract: Edge detection is usually the first step in medical
image processing. However, the difficulty increases when a
conventional kernel-based edge detector is applied to ultrasonic
images with a textural pattern and speckle noise. We designed an
adaptive diffusion filter to remove speckle noise while preserving the
initial edges detected by using a Sobel edge detector. We also propose
a genetic algorithm for edge selection to form complete boundaries of
the detected entities. We designed two fitness functions to evaluate
whether a criterion with a complex edge configuration can render a
better result than a simple criterion such as the strength of gradient.
The edges obtained by using a complex fitness function are thicker and
more fragmented than those obtained by using a simple fitness
function, suggesting that a complex edge selecting scheme is not
necessary for good edge detection in medical ultrasonic images;
instead, a proper noise-smoothing filter is the key.
Abstract: This work presents the hydrogen production from
steam gasification of palm kernel shell (PKS) at 700 oC in the
presence of 5% Ni/BEA and 5% Fe/BEA as catalysts. The steam
gasification was performed in two-staged reactors to evaluate the
effect of calcinations temperature and the steam to biomass ratio on
the product gas composition. The catalytic activity of Ni/BEA
catalyst decreases with increasing calcinations temperatures from 500
to 700 oC. The highest H2 concentration is produced by Fe/BEA
(600) with more than 71 vol%. The catalytic activity of the catalysts
tested is found to correspond to its physicochemical properties. The
optimum range for steam to biomass ratio if found to be between 2 to
4. Excess steam content results in temperature drop in the gasifier
which is undesirable for the gasification reactions.
Abstract: The electrical potentials generated during eye movements and blinks are one of the main sources of artifacts in Electroencephalogram (EEG) recording and can propagate much across the scalp, masking and distorting brain signals. In recent times, signal separation algorithms are used widely for removing artifacts from the observed EEG data. In this paper, a recently introduced signal separation algorithm Mutual Information based Least dependent Component Analysis (MILCA) is employed to separate ocular artifacts from EEG. The aim of MILCA is to minimize the Mutual Information (MI) between the independent components (estimated sources) under a pure rotation. Performance of this algorithm is compared with eleven popular algorithms (Infomax, Extended Infomax, Fast ICA, SOBI, TDSEP, JADE, OGWE, MS-ICA, SHIBBS, Kernel-ICA, and RADICAL) for the actual independence and uniqueness of the estimated source components obtained for different sets of EEG data with ocular artifacts by using a reliable MI Estimator. Results show that MILCA is best in separating the ocular artifacts and EEG and is recommended for further analysis.