Abstract: Camera calibration is an indispensable step for augmented
reality or image guided applications where quantitative information
should be derived from the images. Usually, a camera
calibration is obtained by taking images of a special calibration object
and extracting the image coordinates of projected calibration marks
enabling the calculation of the projection from the 3d world coordinates
to the 2d image coordinates. Thus such a procedure exhibits
typical steps, including feature point localization in the acquired
images, camera model fitting, correction of distortion introduced by
the optics and finally an optimization of the model-s parameters. In
this paper we propose to extend this list by further step concerning
the identification of the optimal subset of images yielding the smallest
overall calibration error. For this, we present a Monte Carlo based
algorithm along with a deterministic extension that automatically
determines the images yielding an optimal calibration. Finally, we
present results proving that the calibration can be significantly
improved by automated image selection.
Abstract: The conjugate gradient optimization algorithm
usually used for nonlinear least squares is presented and is
combined with the modified back propagation algorithm yielding
a new fast training multilayer perceptron (MLP) algorithm
(CGFR/AG). The approaches presented in the paper consist of
three steps: (1) Modification on standard back propagation
algorithm by introducing gain variation term of the activation
function, (2) Calculating the gradient descent on error with
respect to the weights and gains values and (3) the determination
of the new search direction by exploiting the information
calculated by gradient descent in step (2) as well as the previous
search direction. The proposed method improved the training
efficiency of back propagation algorithm by adaptively modifying
the initial search direction. Performance of the proposed method
is demonstrated by comparing to the conjugate gradient algorithm
from neural network toolbox for the chosen benchmark. The
results show that the number of iterations required by the
proposed method to converge is less than 20% of what is required
by the standard conjugate gradient and neural network toolbox
algorithm.
Abstract: We present a new method to reconstruct a temporally
coherent 3D animation from single or multi-view RGB-D video data
using unbiased feature point sampling. Given RGB-D video data, in
form of a 3D point cloud sequence, our method first extracts feature
points using both color and depth information. In the subsequent
steps, these feature points are used to match two 3D point clouds in
consecutive frames independent of their resolution. Our new motion
vectors based dynamic alignement method then fully reconstruct
a spatio-temporally coherent 3D animation. We perform extensive
quantitative validation using novel error functions to analyze the
results. We show that despite the limiting factors of temporal and
spatial noise associated to RGB-D data, it is possible to extract
temporal coherence to faithfully reconstruct a temporally coherent
3D animation from RGB-D video data.
Abstract: The coalescer process is one of the methods for oily water treatment by increasing the oil droplet size in order to enhance the separating velocity and thus effective separation. However, the presence of surfactants in an oily emulsion can limit the obtained mechanisms due to the small oil size related with stabilized emulsion. In this regard, the purpose of this research is to improve the efficiency of the coalescer process for treating the stabilized emulsion. The effects of bed types, bed height, liquid flow rate and stage coalescer (step-bed) on the treatment efficiencies in term of COD values were studied. Note that the treatment efficiency obtained experimentally was estimated by using the COD values and oil droplet size distribution. The study has shown that the plastic media has more effective to attach with oil particles than the stainless one due to their hydrophobic properties. Furthermore, the suitable bed height (3.5 cm) and step bed (3.5 cm with 2 steps) were necessary in order to well obtain the coalescer performance. The application of step bed coalescer process in reactor has provided the higher treatment efficiencies in term of COD removal than those obtained with classical process. The proposed model for predicting the area under curve and thus treatment efficiency, based on the single collector efficiency (ηT) and the attachment efficiency (α), provides relatively a good coincidence between the experimental and predicted values of treatment efficiencies in this study.
Abstract: Web usage mining has become a popular research
area, as a huge amount of data is available online. These data can be
used for several purposes, such as web personalization, web structure
enhancement, web navigation prediction etc. However, the raw log
files are not directly usable; they have to be preprocessed in order to
transform them into a suitable format for different data mining tasks.
One of the key issues in the preprocessing phase is to identify web
users. Identifying users based on web log files is not a
straightforward problem, thus various methods have been developed.
There are several difficulties that have to be overcome, such as client
side caching, changing and shared IP addresses and so on. This paper
presents three different methods for identifying web users. Two of
them are the most commonly used methods in web log mining
systems, whereas the third on is our novel approach that uses a
complex cookie-based method to identify web users. Furthermore we
also take steps towards identifying the individuals behind the
impersonal web users. To demonstrate the efficiency of the new
method we developed an implementation called Web Activity
Tracking (WAT) system that aims at a more precise distinction of
web users based on log data. We present some statistical analysis
created by the WAT on real data about the behavior of the Hungarian
web users and a comprehensive analysis and comparison of the three
methods
Abstract: Speech corpus is one of the major components in a
Speech Processing System where one of the primary requirements
is to recognize an input sample. The quality and details captured
in speech corpus directly affects the precision of recognition. The
current work proposes a platform for speech corpus generation using
an adaptive LMS filter and LPC cepstrum, as a part of an ANN
based Speech Recognition System which is exclusively designed to
recognize isolated numerals of Assamese language- a major language
in the North Eastern part of India. The work focuses on designing an
optimal feature extraction block and a few ANN based cooperative
architectures so that the performance of the Speech Recognition
System can be improved.
Abstract: In this study, a new and fast algorithm for Ascending
Aorta (AscA) and Descending Aorta (DesA) segmentation is
presented using Computed Tomography Angiography images. This
process is quite important especially at the detection of aortic
plaques, aneurysms, calcification or stenosis. The applied method has
been carried out at four steps. At first step, lung segmentation is
achieved. At the second one, Mediastinum Region (MR) is detected
to use in the segmentation. At the third one, images have been
applied optimal threshold and components which are outside of the
MR were removed. Lastly, identifying and segmentation of AscA and
DesA have been carried out. The performance of the applied method
is found quite well for radiologists and it gives enough results to the
surgeries medically.
Abstract: Knowledge management is a process taking any steps
that needed to get the most out of available knowledge resources.
KM involved several steps; capturing the knowledge discovering
new knowledge, sharing the knowledge and applied the knowledge in
the decision making process. In applying the knowledge, it is not
necessary for the individual that use the knowledge to comprehend it
as long as the available knowledge is used in guiding the decision
making and actions. When an expert is called and he provides stepby-
step procedure on how to solve the problems to the caller, the
expert is transferring the knowledge or giving direction to the caller.
And the caller is 'applying' the knowledge by following the
instructions given by the expert. An appropriate mechanism is
needed to ensure effective knowledge transfer which in this case is
by telephone or email. The problem with email and telephone is that
the knowledge is not fully circulated and disseminated to all users. In
this paper, with related experience of local university Help Desk, it is
proposed the usage of Information Technology (IT)to effectively
support the knowledge transfer in the organization. The issues
covered include the existing knowledge, the related works, the
methodology used in defining the knowledge management
requirements as well the overview of the prototype.
Abstract: Motion estimation is the most computationally
intensive part in video processing. Many fast motion estimation
algorithms have been proposed to decrease the computational
complexity by reducing the number of candidate motion vectors.
However, these studies are for fast search algorithms themselves while
almost image and video compressions are operated with software
based. Therefore, the timing constraints for running these motion
estimation algorithms not only challenge for the video codec but also
overwhelm for some of processors. In this paper, the performance of
motion estimation is enhanced by using Intel's Streaming SIMD
Extension 2 (SSE2) technology with Intel Pentium 4 processor.
Abstract: In this paper, a new technique for fast painting with
different colors is presented. The idea of painting relies on applying
masks with different colors to the background. Fast painting is
achieved by applying these masks in the frequency domain instead of
spatial (time) domain. New colors can be generated automatically as a
result from the cross correlation operation. This idea was applied
successfully for faster specific data (face, object, pattern, and code)
detection using neural algorithms. Here, instead of performing cross
correlation between the input input data (e.g., image, or a stream of
sequential data) and the weights of neural networks, the cross
correlation is performed between the colored masks and the
background. Furthermore, this approach is developed to reduce the
computation steps required by the painting operation. The principle of
divide and conquer strategy is applied through background
decomposition. Each background is divided into small in size subbackgrounds
and then each sub-background is processed separately by
using a single faster painting algorithm. Moreover, the fastest painting
is achieved by using parallel processing techniques to paint the
resulting sub-backgrounds using the same number of faster painting
algorithms. In contrast to using only faster painting algorithm, the
speed up ratio is increased with the size of the background when using
faster painting algorithm and background decomposition. Simulation
results show that painting in the frequency domain is faster than that in
the spatial domain.