I will give a trimmed down version of my Masters Presentation. I will outline motivation for the study and jump directly to results and discussion. You have all seen the experiment in action, so I won't bore you with design and implementation issues. Hopefully this will fit in 20 minutes.
The Institute for Computer Research (ICR)
Presents a Colloquium on
A Robust Vision System for Extracting and Recognising
Characters from Natural Scene
by: Dr. Joseph Poon
of: Department of Electronic Engineering
Hong Kong Polytechnic University
Hung Hom, Hong Kong
Date: Wednesday, August 30, 1995
Time: 3:30 pm.
Place: William G. Davis Computer Research Centre, Room 1302
Abstract:
A robust character recognition system will be presented which can
be applied to vehicle licence plate recognition, label identifi-
cation, parcels handling, cargo and aircraft identification. In
general, the practice of extracting and then recognising charac-
ters from scene images require more complicated techniques to
handle 2-D images extracted from 3-D space. Unlike characters
obtained from the scanners, characters extracted from 3-D space
always suffer from a variety of distortion components. For in-
stance, the licence plate characters are usually placed on a more
complex scene which consists of many different objects with a
large range of intensity. Meanwhile, the vehicle may not be
static at the time of taking photo, and so there is some blurring
effect introduced to the image. Besides, the image may not be
captured directly under a camera but at different viewpoints.
Furthermore, the environment cannot be well controlled in such a
way that the character images may be poorly illuminated, occluded
by other objects, or distorted on rainy and smoggy days. There-
fore, from academic and commercial point of views, there is an
urged demand for developing advance character processing tech-
niques suitable for handling scene images.
Biographical Sketch:
Joseph Poon received his PhD degree in 1987 from Sussex Universi-
ty, England. He is currently an Assistant Professor in the
department of Electronic Engineering at the Hong Kong Polytechnic
University. His present research interests include pattern
recognition, image processing and speech signal processing.
Everyone is welcome. Refreshments served.
The Institute for Computer Research (ICR)
Presents a Colloquium on
Coefficient Diagram Method as Applied to the
Attitude Control of Satellite
by: Dr. Shunji Manabe
of: Department of Control Engineering
Tokai University
Kanagawa, Japan
Date: Wednesday, September 6, 1995
Time: 3:30 pm.
Place: William G. Davis Computer Research Centre, Room 1302
Abstract:
The control design theories are roughly classified as Frequency
Response approach (FR), Root-Locus (RL) method, and State Space
approach (SS). The State Space approach is further divided into
Pole Placement (PP), Optimal Control, and H-infinity approach.
The "Coefficient Diagram Method (CDM)" is a control system design
method based on the finding that, when the coefficients of the
characteristic polynomials are plotted on a semi-log scale to
form a "Coefficient Diagram", the shape of the Coefficient Di-
agram is a good measure of stability, response, and robustness of
the system.
CDM looks similar to Pole Placement approach. The difference
lies in that, in CDM, coefficients of the characteristic polyno-
mials is used as the design base, while in PP the location of
poles is used as the design base. Because of this difference,
simple and practical controllers can be easily designed in CDM.
The coefficients of the characteristic polynomial have explicit
mathematical relation with the parameters of controller. Thus,
in CDM, trial-and-error approach like in FR or RL approach (gen-
erally referred as classical control approach) can be efficiently
carried out, keeping a good balance between the necessity of the
system requirements and the complexity of the controller. In
short the CDM is a control system design method which has the
rigor and flexibility of SS approach (generally referred as
modern control) and simplicity and practicality of classical con-
trol.
The attitude control of three-axis-stabilized satellite is real-
ized by three systems, namely, Bias-Momentum-System (BMS),
Controlled-Bias-Momentum-System (CBMS), and Zero-Momentum-System
(ZMS). Since a control law for CBMS was first proposed by
Terasaki in 1967, several control laws are proposed by various
designers. Although each of them was successful in application,
the design philosophy behind them are not clear, because the
problem is beyond the reach of standard FR, RL, or SS, and
designs were made rather in trial-and-error approach.
However, when these designs are analyzed by CDM, it was found
that each controller is fairly similar to the controller which
would be designed by CDM under the given specific circumstances.
Although the controllers look quite different, the resulting
closed-loop systems were found to satisfy the condition of
preferable system from the stand point of CDM.
The presentation is divided into two parts. First, the histori-
cal background, basic theory, and design procedures of CDM are
presented, where "Kessler Canonical Multi-Loop Structure" and the
sufficient condition of stability by Lipatov will be explained.
Also some mentions will be made on how the design can be expedit-
ed using specially designed MATLAB M-File.
Secondly, the control system of CBMS will be analyzed, and vari-
ous control laws in the past have been presented. Also, con-
trollers are designed using CDM, and comparison is made with the
controllers presented in the past. From this comparison, in-
terpretation of each controller is made from the standpoint of
CDM.
Biographical Sketch:
Dr. Shunji Manabe was born in 1930. Dr. Manabe received his B.E.
degree in 1952, from the department of Electrical Engineering,
Tokyo University. He received his M.S. degree in 1954, from the
department of Electrical Engineering, Ohio State University, dur-
ing his stay as a Fulbright Scholarship grantee. He received his
PhD. degree in 1962, from the department of Automatic Control,
Tokyo University.
From 1954 to 1990, Dr. Manabe worked for Mitsubishi Electric Cor-
poration in a varied capacity, specializing in the development of
various control systems in such fields as motor drive, wind tun-
nel drive, train drive, tracking radar control, flight simulator,
power plant simulator, and spacecraft control.
Dr. Manabe joined the department of Control Engineering, Tokyo
University in 1990.
Dr. Manabe's current interest is in the Coefficient Diagram
Method for control system design.
Everyone is welcome. Refreshments served.