Académique Documents
Professionnel Documents
Culture Documents
Smart City
th
Table of
Contents
ICESIT 2016
Page
Program Overview..1
Patrons.4
List of Committee Members..5
Welcome Messages...6
Keynote Speakers..7
Invited Talks15
Industrial Session..26
Technical Program.37
Map..119
Program Overview
August 5, 2016
August 6, 2016
Room M801
09:00-12:00
Room M801
Coee Break
14:30-14:40
Room M801
Room M800
12:00-13:00
13:00-14:30
ICESIT 2016
Registration
17:30-18:00
Room M800
Going to Dinner
Program Overview
ICESIT 2016
Program Overview
ICESIT 2016
ICESIT 2016
Patrons
List of
Committee Members
ICESIT 2016
Advisory Board
Honorary Chairs
General Co-chairs
General Chair
Panuwat Danklang (BUU, Thailand)
(CMU, Thailand)
Message from
TESA President
ICESIT 2016
Werachet Khan-ngern
Thai Embedded System Association, President
ICESIT 2016
The 2016 International Conference on Embedded Systems
and Intelligent Technology
August 5-8, 2016
Keynote Speakers
Keynote Speaker
Watchara Chatwiriya
ICESIT 2016
Dr.Watchara Chatwiriya
Lecturer of Computer Engineering Dep.,
Faculty of Engineering,
King Mongkut's Institute of Technology Ladkrabang
Email: watchara.ch@kmitl.ac.th
Education
2002
1992
1985
Area of Expertise
Teaching Class
Experience
2015 - Present
Board of Director
Bank for Agricultural and Cooperative Agricultural
2011 - Present
ICT Advisor
Office of The Narcotics Control Board (Bangkok)
2008 - Present
Secretary
Thai Embedded System Association
2005 - 2007
Vice Director
Computer Research & Service Center, KMITL
2003 - 2005
Department Chair
Department of Computer Engineering, KMITL
1978 - Present
Lecturer
Department of Computer Engineering, KMITL
Keynote Speaker
Watchara Chatwiriya
ICESIT 2016
Keynote Speaker
Wiroon Sriborrirux
ICESIT 2016
Education
2000 - 2003
1995 - 1998
Bachelor of Computer
Engineering
Research Area
Embedded System
1.
2.
3.
Keynote Speaker
Wiroon Sriborrirux
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
ICESIT 2016
Wiroon Sriborrirux, Panuwat Promsiri, and Apirath Limanee (2014). Multiple Secret Key
Sharing based on Network Coding Technique for Open Cloud DRM Service Provider. IEEE
17th International Conference on Computational Science and Engineering, 953-959.
Wiroon Sriborrirux, Pariwat Leamsumran, Panuwat Dan-klang (2014). Real-time system
for monitoring activity among the elderly using an RF SoC device with triaxial accelerometer
data over a wireless sensor network. The 2014 IEEE MTT-S International Microwave
Workshop Series on RF and Wireless Technologies for Biomedical and Healthcare
Applications (2014)
Wiroon Sriborrirux, Apirath Limmanee, and Sorakrai Kraipui (2013). Unequal Security
Protection: A Unified Framework, Implementation, and Performance Evaluation of Theoretical
and Practical Security. The Tenth International Symposium on Wireless Communication
Systems 2013
Khwanchai Sringiwrai, Nakorn Indra-Payoong, Agachai Sumalee, Pairoj Raothanachonkun,
Wiroon Sriborrirux. RFID-based travel time estimation: development case in Bangkok.
Conference Proceedings of the 15th HKSTS International Conference (Transportation and
Urban Sustainability) (2010)
Wiroon Sriborrirux, S. Kraipui, N. Indra-Prayoong (2009). B-VIS Service-oriented
Middleware for RFID Sensor Network. Singapore 2009 International Conference on Sensor
Networks, Information, and Ubiquitous Computing held in Singapore (2009), 836840.
Wiroon Sriborrirux, Danklang, P., and Indra-Payoong, The Design of RFID Sensor
Network for Bus Fleet Monitoring, In: Proceedings of the 8th International Conference on
Intelligent Transport System Communications, 22-24 October, Phuket, Thailand pp. 103
107, 2008
Warathon, Sumonrat, Wiroon Sriborrirux, Applied Cluster Computing for NP-Hard
Problem, Industrial and Research Projects of Thailand Research Fund, Thailand, 2006
WonJun, Wiroon Sriborrirux, Optimizing A Secure Routing Protocol Using ID-Based
Cryptography in Wireless Mobile Ad Hoc Networks, LNCS (Lecture Note of Computer
Science), Springer, USA, 2004
Wiroon Sriborrirux, Optimizing A Secure Routing Protocol Using ID-Based Cryptography
in Wireless Mobile Ad Hoc Networks, International Conference of the ICOIN2004, Busan,
South Korea, 2004
Wiroon Sriborrirux, DIAMETER strong Security Extensions using Kerberos v5 in Wireless
LAN, Proceeding of the KIPS, Seoul, South Korea, 2002
Wiroon Sriborrirux, Implementation of IPSec over MPLS-based MIP supporting AAA
Server, Proceeding of the KIPS, Seoul, South Korea, 2001
Keynote Speaker
ICESIT 2016
Akkarit Sangpetch
Doctor of Philosophy
2005 - 2010
Master of Science
2001 - 2005
Bachelor of Science
with University Honors
Computer Science
Carnegie Mellon University, USA
Bachelor of Science
with University Honors
Work Experience
2013 - Present
2012 - 2013
Publications
Keynote Speaker
Akkarit Sangpetch
ICESIT 2016
Akkarit Sangpetch, Andrew Turner, Hyong S. Kim, How to Tame Your VMs: An
Automated Control System for Virtualized Services, in Proceedings of the Large
Installations Systems Administration (LISA) Conference, San Jose, CA, USA, 2010.
Andrew Turner, Akkarit Sangpetch, Hyong S. Kim, Empirical Virtual Machine Models for
Performance Guarantees, in Proceedings of the Large Installations Systems
Administration (LISA) Conference, San Jose, CA, USA, 2010.
Research
2015 - 2016
2014 - 2015
2010 - 1013
2009 - 2010
2008 - 2009
2006 - 2007
2005 - 2006
Keynote Speaker
Akkarit Sangpetch
ICESIT 2016
Teaching Experience
B.Eng. in Computer Engineering: Cloud Computing, Operating Systems, Unix Operating
System, Object Oriented Analysis & Design
M.Eng. in Computer Engineering: Special Topics in Software Technology
Guest Speaker
2014-2016: Trainer for Thailands International Olympiads in Informatics (IOI) contestants at
The Institute for the Promotion of Teaching Science and Technology
December 2, 2015: Guest speaker Cloud computing direction for 2016 in Leveraging the
Cloud seminar, held by KSC Commercial Internet Co. Ltd.
August 23, 2013: Guest speaker Cloud Technology and related research in seminar for
M.S. in Computer Science students, Kasetsart University
Academic Services
2014-2016: IOI Contestant selection subcommittee member, Institute for the Promotion of
Teaching Science and Technology, Ministry of Education
2014-2016: Information technology project screening subcommittee member, King
Mongkuts Institute of Technology Ladkrabang
2015-2016: Computer innovation engineering curriculum development committee member,
Faculty of Engineering, King Mongkuts Institute of Technology Ladkrabang
2014-2015: Python programming language textbook for public schools editorial board
member, Institute for the Promotion of Teaching Science and Technology, Ministry of
Education
2014-2015: Computer engineering graduate program management committee member,
Faculty of Engineering, King Mongkuts Institute of Technology Ladkrabang
2013-Present: Information Technology Advisor, Netbay Co. Ltd.
2013-Present: Cloud Technical Advisor, Internet Thailand Public Co. Ltd.
ICESIT 2016
The 2016 International Conference on Embedded Systems
and Intelligent Technology
August 5-8, 2016
Invited Talks
Abstract
A new simple chaotic system with both dynamic and jerk forms is presented through the use
of an inherent hyperbolic sine function in anti-parallel diode configuration. The circuit
reveals a complex back-to-back twisted chaotic attractor which has never been found in other
circuit configuration. Dynamic properties are described through an existence of attractor,
Equilibria, Jacobian matrices, bifurcations, and time-domain chaotic waveforms. The
proposed chaotic oscillator is applied for secured RF transmission through synchronization
between transmitter and receiver.
Abstract
Pulsed Electromagnetic Field (PEMF) technologies have been useful in terms of adjunctive
therapy for the treatment of both delayed-union fractures and chronic wounds. Such a PEMF
is a relatively simple device, which employs an external, non-invasive PEMF to generate
shorts bursts of electrical current in injured tissue without producing heat or interfering with
nerve or muscle function. This paper aims to review major scientific breakthroughs and
current understanding of the mechanism of action of PEMF therapy. A literature review is
conducted, involving mechanism of action and biologic and clinical studies of PEMF. This
review shows that surgeons have a powerful tool with no known side effects for the
adjunctive, noninvasive, nonpharmacologic management of postoperative pain and edema.
Recent rapid advancement of portable and economical PEMF devices have revealed that the
PEMF devices are significant to the surgeon for improving outcomes of decreased pain and
swelling following injury or surgery.
Abstract
As the processor-memory performance gap increased, the multilevel cache is introduced. For
example, the Intel Pentium processor has level 1 on-chip cache and level 2 off-chip cache
implemented; the L2 cache moved up to be on-chip in the Intel Pentium Pro. With the advent of
multicore processors, several private and shared levels of cache are used, i.e., the Intel
Core Xeon X5550 has four cores and three layers of cache: each core has private L1 and L2
cache; the Last-level cache (LLC) is shared among four cores. Sharing the LLC can reduce the
number of duplicate copies when many cores are running identical applications.
This work concerns with non-inclusion cache which a datum can have multiple copies like an
inclusive cache but when a datum is discarded from the lower level cache, the copies in the upper
levels are not required to be discarded too. When the datum moves up from lower level to higher
level, the exclusive cache will force deletion of the copy in lower level while the non-inclusive
cache will allow leaving the copy in lower level, avoid wasting any blocks.
Many workloads today are multimedia applications which load millions of blocks of data and use
them only once. Some data are placed in cache and never reused again until they are evicted
from cache. This is especially true in the LLC, since exploitation of temporal locality in high
level cache means an inversion of temporal locality on LLC; in other words, data that is accessed
frequently will always hit on level 1 caches, thus remain unused on levels 2 and 3. Study shows
that numerous data blocks are allocated to the last-level cache and never reused or accessed
again. Those blocks that are placed in the LLC and never accessed should never be allocated on
cache to waste precious cache space. The method of not allocating some data to cache is called
cache bypassing.
Most bypass techniques have relied on ad hoc methods such as counters and tables which cannot
tackle the complexity of multicore workloads. In this work, we propose an alternative method to
predict cache bypassing using Support Vector Machine (SVM) models. Based on access traces
obtained from representative benchmarks running on the Multi2Sim simulator, supervised SVM
training was performed in order to obtain a bypass prediction model suitable for LLC in multicore processors. The SVM outputs bypassing classifiers which are integrated on the simulator to
quantify LLC performance improvements. Results show that, with appropriate parameters and
kernel functions, SVM is capable of generating bypassing models which improve LLC
performance on multicore processors, achieving an average 5.34% hit rate improvement across
SPLASH2 benchmark combinations.
keywords: machine learning, last level cache, cache bypassing
Abstract: Today more functionality is embedded into smaller devices. We need to concern
about security aspect of these devices. An embedded system is a computer system with a
dedicated function within a larger system, often with real-time computing constraints.
Embedded systems usually are designed to perform some specific task, rather than be a generalpurpose computer for multiple tasks. Embedded systems control many devices in common use
today. Most of all microprocessors are manufactured as components of embedded systems. For
example, intelligent techniques can be designed to manage power consumption of embedded
systems, portable devices such as digital watch, city traffic light, and hybrid vehicles.We discuss
some of the security concerns of embedded systems.
Evolution-based method
for clustering of high dimensional data streams
Kitsana Waiyamai *
Department of Computer Engineering, Faculty of Engineering, Kasetsart University, Bangkok, Thailand
Abstract
In recent years, clustering of data streams has become a research topic of growing interest. The streams clustering processes
data in a single pass and summarizes them in real-time, while using limited resources. Streams clustering algorithms that support
the monitoring and the change detection of clustering structures are called evolution-based stream clustering method. Many
techniques have been proposed for clustering data streams. However, very few have proposed to monitor and detect change of the
evolving clustering structures. E-Stream is an evolution-based stream clustering method that supports various types of clustering
structure evolution which are appearance, disappearance, self-evolution, merge and split. In this work, three different techniques
are proposed and evaluated in order to improve the performance of E-Stream in terms of both clustering quality and execution
time. First, to deal with high-dimensional data streams, a projected clustering mechanism is proposed to determine appropriated
subset of dimensions for each cluster. Second, the idea is to exploit background or domain expert knowledge as constraint for
improving the clustering process. Two types of instance-level constraints, Must-link and Cannot-link, are introduced and
integrated into E-Stream. Third, an online-offline-phases model is added into E-Stream. In the online phase, E-Stream is able to
support different evolving clustering structure types. In the off-line phase, the affinity-propagation clustering is adopted to deliver
the final clustering without any need for user intervention.
Keywords: stream clustering; affinity propagation clustering; off-line mode clustering; clustering structure change
detection, projected clustering, constraints-based clustering; semi-supervised learning
ABSTRACT:
This research has proposed an optimization algorithm-integrated adaptive array antenna
(AAA) for the 2.4-2.5GHz wireless communication system. In the realization of the
adaptive antenna system, the fruit fly optimization algorithm (FOA) and modified fruit
fly optimization algorithm (MFOA) were deployed in the processing unit, respectively. In
this research work, simulations and experiments were carried out with the four-element
AAAs of two configurations, i.e. the linear and planar array configurations. The
simulation and experimental results revealed that the MFOA algorithmic scheme could
determine the direction of the maximum arrival signal in an efficient and accurate manner
and also was capable of manipulating the radiation pattern in the desired direction.
At the present, eye gaze tracking system is gaining interest as a powerful system for
helping a disable person. Eye-gaze tracking is the process of measuring the eye movement
during look at different positions on the screen and was captured by the camera. The eye
gaze image was computed and transformed to the screen positions.
Therefore, how to communicate with others people. It is more challenging to develop
the modern technology to support for diseases peoples. In addition, there are many
application require eye-gaze technology to interface between the human and the computer.
An example of the eye-gaze system application such as eye-gaze tracking control a wheel
chair, an embedded system for a Surgical Microscope. If the eye-gaze tracking system can be
detecting quickly and interpreting for the eye-gaze positions on the screen more accurate, the
efficiency of the eye-gaze system will getting higher. The user can using the eye-gaze to control
the computer more accurate. It is more challenge to be using the human eye as an input
replace the keyboard or traditional input.
The higher precision of eye-gaze tracking system is required for the eye-gaze tracking
system. The eye-gaze distance estimations is conducted to computing the distant between
the users eye and the computer screen. Previously, we proposed the eye-gaze distance
estimation using the iris region eigenvalues for user-independent eye-gaze distance estimation
with using a single camera for detection the eye-gaze distance. The trained system was used
only from one eye-gaze image information and used only single camera. The test distance
estimation is 60cm to 80 cm. The experimental results obtained by linear, logarithmic, and
power regressions were 95.98%, 95.33% and 96.37% accuracy.
This research presents a new design method and a new filter structure for the maximallyflat variable fractional-delay (VFD) FIR filter. Since the design method and filter structure are
formulated from the so-called discrete Pascal transform (DPT) and its Pascal interpolation, the
resulting VFD filter is called Pascal VFD filter. The kth-order Pascal polynomial in the Pascal
VFD filter is used for fitting (k+1) data points. The Pascal VFD filter is a real-time VFD filter
that consists of two sections, which can be realized into the front-end and the back-end sections.
The front-end section contains multiplierless digital filters, while the number of multiplications
in the back-end section is a linear function of the order k. Since the proposed Pascal VFD filter
structure can adjust the delay parameter online, it is more suitable for real-time applications.
Moreover, the proposed Pascal VFD filter structure is also divided into two types as the cascadetype and parallel-type Pascal VFD filter based on choice of front-end section. Both two types of
Pascal VFD filter use the same number of multiplications. Consequently, Pascal VFD filter can
give the low-complexity filter structure including no transient error during online adjustment of
delay parameter.
Abstract
This research presents a design of a hybrid rectangular and circular waveguide resonator.
The resonator is designed specifically for the purpose of granular dielectric
measurements, which resonates at 2.45 GHz. The electric field distributions of the
proposed resonator were compared against ordinary resonators, and illustrated using a
3D computer simulation of isosurfaces and their contours and 3D projections. In the
simulation the properties of samples (paddies and soils) are loaded into each design, in
order to observe the electrical field distributions of each sample. Next, the probability
density functions (PDF) was used to determine the probability of electric field
distributions within each type of resonator. Each designed resonator was created in
order to process real experiment circumstances. The experiments for each resonator
were processed using 20 paddy samples and 20 soil samples, in order to replicate real
circumstances in an efficient manner. The percentages of moisture content (%MC) from
each design were compared against %MC obtained by dry basis methodology, used to
find errors. The results showed that the proposed resonator has less error than the
rectangular and circular resonators which are 16.546%, 39.006% and 27.732%
respectively for paddy samples, and 17.088%, 32.583% and 23.638% respectively for
soil samples. This demonstrates that the proposed resonator creates better results than
the ordinary resonators, although the sample is changed.
Keywords:
AbstractSmart electric vehicles (EVs)must composes of three main functions including detection,
communication and decision. Car revolution has been started from steam power - electric power - internal
combustion engine (ICE). Nowadays, it seems to transition from ICE era to EV era. It may take some transition
periods due to the storage energy density aspect. Lithium batteries are common used in EVs which the energy
density (kWh/liter) is one ninth of gasoline. The energy density of the Lithium battery is one of the constraint of
EV driving range. Charging and driving are two EV main performance. While the connecting is a charming of
digital era for connected car. The EV charging can be categorized into wire and wireless charging in term of
connection or be categorized into normal charging and fast charging. Charging is an issue that EV can be a
drawback. Driving range of the EV depends on the energy storage of lithium batteries. Wireless power charging
(WPC) is one of option for the energy charging convenience. The pricing of the battery can be dominated on
the whole EV priceif the large battery is applied like in Tesla car. The connecting EV is an interesting issue not
only the IOT issue, but also can be the power backup to the micro grid to maintain the stability of the electric
grid power supply. Automate driving with or without man is possible which related standards and regulations
are required.
KeywordsSmart
electric
vehicles,
charging,
driving,
connected
car,
EV
ICESIT 2016
The 2016 International Conference on Embedded Systems
and Intelligent Technology
August 5-8, 2016
Industrial Session
ICESIT 2016
ICESIT 2016
Abstract: This is an Embedded System has been developed for new cars and used cars, the
purpose is to customize the functionality of the original ECU module (Piggy-back ECU) that
setup for engine lifetime and pollution control and specific for , the system is suitable for
vehicles modified fuel system , such as LPG or for E85.
This embedded system also developed for modified car for competition (Stand-Alone ECU),
its will take over control the engine and let technicians or engineers adjust by using computer
laptop connects with the system for maximize efficiency and power output.
ICESIT 2016
Sao Denki
Abstract Air Conditioner plays important roles for human comfort. Saijo Denki developed
Internet & Smart phone connected air conditioner using own developed IoT Solution. The air
conditioner can be controlled from anywhere in the world, report its status, intelligently
analyze and inform customers more than 30 error codes. Moreover, the air conditioner can
also report power consumption of air conditioner to customer's smart phone. Integrated with
GPS technology, Saijo Denki IoT air conditioner can send notification to end-user when
approaching home in approx. 3 km. So, customer can switch on AC before reaching &
getting more comfort during extremely hot or extremely cold climate.
ICESIT 2016
Introduction
ICESIT 2016
AbstractThere are a lot of reasons that vending machine will become more and more
popular in Thailand. One of the reasons is the global recession, when people cant make
money easily; they tend to find income from selling commodities. The first successor is the
mobile topup vending machine. Its success has raised a lot of company in this market. And
when the market is becoming full, these companies ask Whats next? and many kind of
vending machine are the answers. In the same time that IoT rise, it helps developers to
develop their systems smarter and in easier way.
KeywordsVending machine, IoT, mobiletopup
ICESIT 2016
Abstract: According to the business influences on our Company caused by the world's
megatrends of global warming, labor shortage, smart automation and growth of elder
population worldwide, including Thailand, SCG started to set up research facilities called
SCG Housing Technology Center in 2008 at National Science Park area. Three research
areas of smart-eco-care are focused. Six years later, we selected several prominent
technologies to construct a unique prototype house named The NEST: The Next Eco
Sustainable Technology. Smart automation technology was integrated with many other
products to enhance the housing performance, safety, security and convenience functions of
The Nest. Active AIRflowTM was developed to increase the quality of air ventilation of the
house and allow the air to be ventilated throughroofing tile ventilators by using smart logic
control with cloud operating platform. Automatic night light was developed for elderly people
who need special care during night time when they wake for toilet and it enabled them to
return to their bed without any injuries and better continue their sound sleep under the warm
light. Embedded logic protocol was developed in order to achieve smoothly dimmable
lighting while individual user could enjoy any activities as long as they want. Smart lighting
control switches were also one of the products being installed in The NEST mainly for
convenienceproposes. Having been testing all functions in The NEST for more than 2 years, it
can be summarized from what we found that the majority of our customers prefer to invest
their money on safety and security rather than on their convenience.
Keywords: Smart living, Automatic night light, Roof ventilation, Elder care
ICESIT 2016
ICESIT 2016
ICESIT 2016
Patchpon Asawachiwantorngul
Crafity Studio Co., Ltd. Bangkok, Thailand
Email: patchpon@crafitystudio.com
ICESIT 2016
Introduction According to our business, BAESLab Company Limited has been the spin-o
company from Embedded Systems Laboratory by Mr.Thanong Chotisorayuth and SE-Education
Public Co. Ltd since 2013. Our mission is to be the new generation of design house in Thailand.
By our strengths and knowledges in the Embedded systems technologies, we develop the
Smartcity solutions joins hand with Saensuk municipality and other strategic partners. For the first
phase of this project, it is designed for smart healthcare under concept of Elderly care at home.
So, The elderly peoples will have their activity monitored by their families or nurses via the smart
wristband or smart necklace including the fall detection function. In the other hand, we also
develop the Advance Service Framework to compatible with the various operating systems
especially Linux over the serveral IoT Gateway devices.
ICESIT 2016
The 2016 International Conference on Embedded Systems
and Intelligent Technology
August 5-8, 2016
Technical Program
acquire high quality depth maps [2]. To achieve this goal, first
we propose a moving camera array. This camera systems
captures multi-view color videos at 30 frame per second and
color resolution is 1280720. To capture the moving object, our
proposed multi-camera array can move left and right sides.
Second, we propose a depth map refinement method to improve
a virtual views quality. The proposed algorithms is based on
noise aware filter for depth upsampling (NAFDU) [3]. Also, we
consider reliability of color and depth pixels to solve an edge
blurring problem. Third, we propose a virtual view synthesis
method in the moving multi-camera array. In our method, we
synthesis virtual viewpoint images using both spatial and
temporal information. Once the disparity maps are determined
from left and right image sequences, we are able to synthesis a
virtual view flexibly. After that, we search the reference image
using the color differences between the virtual view and
candidate images. Then we divide the moving and static areas
differences virtual view image and reference image. Finally, we
obtain the improved virtual view images using reference
information.
I. INTRODUCTION
Virtual view synthesis is one of the most important
techniques to make high quality videos in the free viewpoint
television and three-dimensional (3D) video. Unfortunately,
efficiently making virtual view of complex real world locations
is a challenging task. Depth image-based rendering (DIBR) is
generally used to synthesize a virtual view in free viewpoint
television (FTV) and 3D video [1].
In the previous works, most efforts were focused on static
multi-camera system. Many static multi-camera systems were
presented to capture the scene and generate 3D video contents.
Most of them are composed of fixed multiple color cameras
with parallel arrangement. In this case, if an object moves
widely, the number of camera should be increased. As a result,
camera setup and computation cost of the algorithm will be
increased. Alternatively, we can reduce the number of cameras
by increasing the camera interval. However, in this case, only
sparse information can be used in the view synthesis, so that the
quality of the generated free-viewpoint image significantly
degrades and it is difficult to expect high quality of the 3D video.
The contributions of this paper include virtual view
synthesis method in the moving multi-camera array, and
of pixels in the filter kernel. Also, we can get the mean value in
the filter kernel using (2).
1
(2)
The standard deviation is a measure of how spreads out
pixel values are. In addition, it can represent the dispersion of a
set of pixel values from the filter kernel. So, we can decide
which region is the edge by using standard deviation. Our
proposed method use the adaptive weighting functions based on
edge information. Blending function is given by Eq. (3).
(1)
1
1
,x
(3)
max
(4)
(7)
(8)
where and are the left and the right images, respectively. x
and x are pixel positions, d is disparity vector and is the ratio
of baseline. In the moving multiple camera array, we can
synthesize virtual viewpoint images using conventional linear
interpolation method.
IV. CONCLUSION
In an ideal parallel fixed camera array, cameras are located on a
certain line which is called the baseline. Also, all the cameras have the
same distance to the adjacent cameras. In this case, range of virtual
viewpoint image generation is limited. To solve this problem, our
proposed method using the moving multiple camera array. We apply
the semi global matching and multilateral filtering to get the high
quality disparity map and synthesis virtual viewpoint images using
both spatial and temporal information. As shown in the experimental
results, we can obtain a high quality of the virtual viewpoint images
for moving camera array. For most parts of the comparison, our
proposed method showed the better results with the results of the
conventional methods.
V. ACKNOWLEDGMENT
This research was supported by Basic Science Research Program
through the National Research Foundation of Korea(NRF) funded by
the Ministry of Science, ICT & Future Planning(2011- 0030079)
VI. REFERENCES
Watchara Chatwiriya
I. INTRODUCTION
IoT is more practical in many application areas because an
increasing number of smart devices and ease of Internet
accessibility. Even though the concept of IoT is defined, but
the communication protocols to be used with are still in the
developing and competing to be standard [12].
One of the proposed protocols is MQTT, a lightweight
protocol with many advantages [1]. MQTT use publishsubscribe pattern in which the server (This server is called a
broker) uses with message intermediate [1,2]. In this pattern, a
broker will receive the data from devices, and forward to
connected clients, as listed in topic of devices. The broker will
forward data to that client only if the client subscribes common
topics.
In MQTT standard [1], QoS is defined as the service level
agreed between sender and receiver for message sending, with
the guarantees of delivering messages. Actually, the sender has
to keep sending until it gets the response, to meet QoS
conditions. With restrained bandwidth and noisy environment,
the waiting time for this pattern is unpredictable and
ineffective.
In this paper, we proposed a solution to solve this problem,
assumed there is nearby communication devices or other
communication channel. HTTP and ARP Protocol can apply by
sending data to nearby devices and retransmit data to a broker
instead of ordinary sender using.
The rest of the paper is organized as follows. Section 2
describes background knowledge. Section 3 discusses about the
concept of the proposed solution. Section 4 explains about
implementation. Section 5 describes the experiment. Lastly,
section 6 concludes the work.
II. BACKGROUND
IoT communication protocols are developed to meet
various requirements such as support heterogeneous hardware,
capable of data management, support application developers,
extend the different platforms for the formation of ecosystems
and the availability of dedicated marketplaces to the IoT [11].
One of the popular protocols is MQTT. The communication
model in MQTT can be described as sending messages
between brokers and senders, with QoS mechanism to prevent
disruption in the transmission. The QoS is described as three
levels as the following [13]:
QoS level 0 sends the message only once following the
message distribution flow, and does not check whether
the message arrived to its destination. Therefore, in case
of sizable messages. It is possible that the message will
be lost when any kind of loss comes in the way.
QoS Level 1 sends the message at least once and checks
the delivery status of the message by using the status
check message, PUBACK. However, when PUBACK
is lost, it is possible that the server will send the same
message twice, since it has no confirmation of the
message being delivered.
QoS level 2 passes the message through exactly once
utilizing the 4-way handshake. It is not possible to have
a message loss in this level, but due to the complicated
process of 4-way handshake, it is possible to have
relatively longer end-to-end delays.
IV. IMPLEMENTATION
In this section, we illustrate the network diagram of system
implementation as Fig. 5. If one of the network is not working,
primary sender can transmit messages to nearby devices which
will forward the messages to the broker server.
if result == Success:
returnSuccessResultToPrimarySender()
stopAllProcess()
else:
else:
returnFailResultToPrimarySender()
findNearbyDevice()
Fig. 6. Show pseudocode about condition of MQTT sending process
V. EXPERIMENT
In this section, we show the experimental setup and the
result. By we set up the system as the following:
Open sources which we choose in this experiment are
Mosquitto as a broker server and Paho as clients.
There are 1 broker, 1 receiver, 1 primary sender, only 1
nearby device which can forward messages to the
broker and 10 other nearby devices in the network.
allNearbyDeviceIp = ARP()
Fig. 7. Showpseudocode about store IP Address to array list
Fig. 10. Result when network congestion has occurred for 10 seconds
QoS 1
QoS 2
Proposed
Solution
10
10.81
11.01
14.88
Average
quantity of
forwarded
devices (unit)
7.2
20
20.91
21.01
14.83
7.0
30
30.91
30.81
14.93
7.4
Network
congestion
time (second)
Fig. 11. Result when when network congestion has occurred for 20 seconds
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Fig. 12. Result when when network congestion has occurred for 30 seconds
Advancing
open
standard
for
the
information
society,
https://www.oasisopen.org/
U. Hunkeler, H.L. Truong and A. Standford-Clark, "MQTT-S - A
publish/subscribe protocol for Wireless Sensor Networks," 3rd
Communication Systems Software and Middleware and Workshops,
Bangalore, IEEE, 2008, pp. 791-198.
RFC Editor, https://www.rfc-editor.org/
Y. Qin, Y. Zhou, and B. Liu, "A High Performance ARP Lookup
System for Gigabit Ethernet," 2nd Communications and Networking in
China, Shanghai, IEEE, 2007 pp. 82-86.
V. Gupta, R. Goldman, and P. Udupi, "A Network Architecture for the
Web of Things," 2nd International Workshop on Web of Things, San
Francisco, ACM, 2011, article. 3.
] C. Zhang, C. Cheng, and Y. Ji, "Architecture Design for Social Web of
Things," 1st International Workshop on Context Discovery and Data
Mining, Beijing, ACM, 2012 article. 3.
B. Aziz, "A Formal Model and Analysis of the MQ Telemetry Transport
Protocol," 9th International Conference on Availability, Reliability and
Security, Fribourg, IEEE, 2014 pp. 59-68.
[8]
PROPOSED METHOD
In this paper, the proposed method is used to measure infocused regions and combine these regions to all-focused
image as shown in Fig. 1. The following techniques are
involved in the proposed methods implementation.
I.
INTRODUCTION
1) Approximation coefficient
Approximation coefficient is the source image at the
coarser resolution level, which can be considered as a
smoothed and subsampling version of the source image. Most
information of the source image is kept in the approximation
coefficient. Then, we utilize the spatial frequency (SF), which
indicates the overall active level in an image and originated
from the human visual system, on the approximation
coefficient. The SF is expressed as:
(1)
(2)
(5)
(3)
III.
EXPERIMENT RESULTS
2) Detail coefficients
The detail coefficients contain the detailed information of
an image, which usually have large values correspond to sharp
intensity changes and preserve salient information in the image.
Moreover, The energy of the detail coefficients of an infocused image is much larger than that of a blurred image.
Then, we apply sum-modified-Laplacian (SML) to the detail
coefficients. It can effectively represent the salient features and
sharp edge of the image. The SML is expressed as:
Images
Figure 4.
Figure 5.
Criteria
Pixel
Avg.
DWT
SF
Proposed
Method
MI
4.147
3.943
4.453
4.608
PSNR(dB)
26.994
24.357
28.232
28.310
MI
7.155
7.598
7.113
7.818
PSNR(dB)
24.542
21.324
24.517
24.831
IV.
Figure 4.
CONCLUSIONS
[2]
[3]
[4]
[5]
Figure 5.
[6]
[7]
[8]
I. MODEL-BASED DEVELOPMENT
Model-based development (or MBD in short) [1] is an
approach for developing embedded software using system
models as development artifacts that link from requirement
phase to testing phase. Representing architecture and
behaviors as models, developers can apply simulation tools to
validate and verify system operations early in the development
process. Consequently, several industrial sectors acknowledge
the necessity and advantages of MBD approaches in their
software projects especially those with safety critical
requirements. For example, automotive software development
has migrated to the MBD approach in order to cope with
expanding scope and complexity of software functions.
The MBD approach differs from the tradition approach in
several aspects. The flow of conventional software
development starts by analyzing software architecture and
functions from text-based documents, writing code to fulfill
features, and testing code according to specifications in the
original documents. Such flow demands engineers and
developers with different technical backgrounds and skill sets
to work together. While the MBD approach starts by modeling
functions and behaviors from specifications stated in the textbased documents, then models will be used and refined
throughout development activities. Using simulation and code
generation tools, simulated results and execution image can be
obtained for verification and hardware deployment,
respectively. Hence, timing behavior and computation
accuracy can be evaluated before launching a field trial.
At present, the main obstacle in applying the MBD
approach is differences in development tools, development
A. Learning Objective
Embedded software development usually relies on the Vmodel development process [4] which abstracts development
activities into the specification stream (requirement, design)
and the verification stream (testing) with coding as the joint
activity between both streams. The V-model can be extended
for the MBD approach by considering modeling and
simulation techniques for each phase of development.
1.
2.
3.
4.
5.
TABLE I.
MILS
LEARNING SCOPE
SILS
2.
3.
4.
Hardware deployment.
HILS
Simulation
Deployment
B. Problem Setting
Corresponding to the automotive industry, the problem is
selected to be the electronic throttle control (ETC). The ETC
problem offers several advantages with respect to its
complexity layers. The simplest layer is to design a feedback
controller for the angular position control of DC motor. The
design problem becomes more complicated by considering
operating modes, e.g. failure modes and cruise control.
RCP
Controller design
1.
PILS
Plant modeling
C. Hardware Platform
The hardware platform for the MBD education consists of
two processor boards and an I/O board as shown in Fig. 4. The
processor board is the STM32F4 Discovery board [5]
equipped with STM32F407VG, 32-bit ARM Cortex-M4
microcontroller running at 168MHz. This board was chosen
for its availability of code generation tool, that is the Waijung
library [6].
TABLE II.
Day
1st
2nd
3rd
4th
5th
Activity
Topics
II
Workshop
Assignment
Assignment
Assignment
Assignment
Hardware simulator
Assignment
Lecture
Workshop
Lecture
2.
3.
4.
5.
POINT SUMMARY
Avg
Max
1st
3.6
36
8.1
81
2nd
15.6
63
19.9
79
Signal generation
11.6
47
22.7
91
4th
8.7
29
29
97
5th
Presentation
7.2
71
8.9
89
rd
Topics
Total
46.8
88.6
Acknowledgment
The author would like to express our deep gratitude to
event sponsors and Top Gun Rally technical committees for
their financial support and knowledge contribution.
References
[1]
[2]
[3]
[4]
[5]
IV. DISCUSSION
The 9th TESA Top Gun Rally was the first academic event
to target the MBD education in Thailand. Due to the time
constraint and technical gaps, the learning outcome as
expressed by 46.8 average points should be improved by
[6]
I. INTRODUCTION
Disparity map generation from stereo images is one of the
fundamental process for three-dimensional reconstruction and
has been researched into various methods. One of these methods
is called as stereo matching, which is estimating the depth of
projected view from stereo images of different views. Generally,
it is divided into two classes; one is local method and another is
global method.
Local methods on stereo matching are estimating the
disparity with comparison of the matching cost from views on
the right and the left. It is simple to implement and fast to be
suitable for real-time applications. For calculating matching
costs, the size of the matching window must be determined at
first, which has an effect on the quality of disparity map. The
disparity map from using local methods usually shows low
accuracy than global-based methods.
Global methods on stereo matching are based on Markov
random fields and estimating proper disparity values by
minimizing predefined energy functions. Energy function can be
minimized by using optimization algorithms such as belief
propagation, graph cut and dynamic programming technique.
Energy optimization process takes the relationship between the
entire pixels in the image into account, so it is slower than local
methods.
In this paper, we applied gradient stereo images with census
transform to improve the accuracy of disparity values on the area
of depth discontinuities. Gradient images contain the edge
information of original stereo images and it is related to the
depth discontinuity regions. Basically we exploited belief
propagation with intensity-based census transform to calculate
1
mij = min[( , ) + ( ) + () ,
] (1)
( , ) = min(| |, )
(2)
bi ( ) = ( ) + () ,
(3)
(4)
1, > + [, ]
(, + [, ]) = {
0,
(5)
IV. CONCLUSION
[2]
[3]
[4]
I.
INTRODUCTION
II.
LIBRARY DESIGN
Block
LIBRARY STRUCTURE
Function
Toggle status of digital output pin
for processing time measurement
Convert analog signal to digital
value at 1024-Hz sampling rate and
12-bit resolution.
Read
data
from
3-axis
accelerometer, 3-axis gyroscope and
3-axis magnetometer in an IMU
module at 256-Hz sampling rate.
Time (second) =
EVALUATION
(1)
[2]
[3]
[4]
Fig. 6. Processing Time of Model in Figure 5.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
= |( + 1) ()|, (1)
= 1 = 1 = 1
where () represents height, top position, and bottom
position of silhouette, height is distance between bottom
and top positions at frame , is number of class, is
number of template, is number of frame, is class, is
template, and is frame.
Then we separate occlusion and non-occlusion by
comparing history of height, top position, and bottom position
in testing silhouette image sequence with its threshold. If
history is more than threshold that frame is in
occlusion period. The history of silhouette height, the top
silhouette position, and the bottom silhouette position is
calculated as follows.
= 2
(2)
1(( + 1) ()),
where x1 is the first frame of changing height direction,
and x2 is the last frame of changing height direction.
Table 1 Adaptive reference position rule.
Rule
R1
R2
R3
R4
R5
R6
R7
R8
R9
R10
Ip1
1
1
1
1
1
1
1
1
1
0
Ip2
1
1
1
0
0
0
0
0
0
x
Input
Ip3
Ip4
0
1
0
0
0
0
1
1
1
0
1
0
0
1
0
0
0
0
x
x
Ip5
0
1
0
0
1
0
0
1
0
x
Ip6
0
0
1
0
0
1
0
0
1
x
Output
Op1
Op2
O11
0
O12
0
O13
0
0
O21
0
O22
0
O23
O14
O24
O15
O25
O16
O26
0
0
1
2 1 +1)
2
=
(),
1
(3)
2
=
() ,
1
2 1 +1)
where () is the height of silhouette frame in nonocclusion period which is before occlusion period.
= (
1
2 1 +1)
= ( + ),
(5)
2
=
() ,
1
(6)
Meaning
Occlusion state of height of silhouette image
Occlusion state of top position of silhouette
image
Occlusion state of bottom position of silhouette
image
Occlusion period at first frame of silhouette
image sequence
Occlusion period at middle of silhouette image
sequence
Occlusion period at last frame of silhouette
image sequence
Compensating top reference position ()
Compensating bottom reference position ()
() = () +
() = () +
()
()
=
+
()
() =
()
=
()
= ()
()
= ()
() = () +
()
() =
()
= (
1
2 1 +1)
2
=
() ,
1
= (
(4)
(7)
()
= (
(8)
2
=
() ,
1
(9)
1
2 1 +1)
()
where
is the bottom position of silhouette frame
in non-occlusion period which is after occlusion period.
2
=
() ,
1
(10)
where
is the bottom position of silhouette frame
in non-occlusion period which is before occlusion period.
(
+
)
(11)
= ,
(12)
, = , ,
(13)
= ,
(14)
= , ,
= 1 = 1
( + )
= 2 ,
1
2 1 +1)
(15)
position is occluded, but also in frame which both the top and
the bottom positions are occluded. Moreover, we can use
information of GEI, which is constructed from the adaptive
reference position, in advancing gait identification as
separating the occluded GEI part and the non-occluded GEI
part as Nangtin et al. [2].
Table 3 Recognition rate.
Dataset
Sequence
EEPIT
DuP
DlP
LfUp
LfLp
MfUp
MfLp
RfUp
RfLp
CASIA
Normalization method
NonAdaptive
adaptive
66.67
69.67
20.00
26.53
33.60
40.52
56.91
59.85
23.85
30.53
48.24
53.78
32.52
35.24
67.48
69.52
Different
3.00
6.53
6.92
2.94
6.68
5.54
2.72
2.04
6. CONCLUSION
Fig. 3 An example of type of occlusion on all scenes in
EEPIT dataset [5].
I. INTRODUCTION
III. PROPOSED
ROPOSE METHOD
A. IC panel detection
We propose a method to inspect the IC panel from the
input image in this topic which can divide in to three phases.
1) Convert RGB to HSL
The RGB image was contain image size 32 bits, so we need
to decrease image size to 8 bits. The method of HSL
conversion is the properly for clarify, this paper choose and
coverts RGB image to HSL.
2) Set reference pattern
After convert image to HSL, we need to set reference
pattern for start from this point to search panel edge in image,
this paper we propose to use ring notch as reference pattern.
3) Set panel edge
After get reference pattern, In this paper use edge
detection to set region of interest to find panel edge of 4 sides
by search from bright to dark as Fig. 3. After set panel edge
complete. We will get the region of interest and this function
we will use panel edge position for setting offset pixel to find
start position
C(i, j )
w( x, y) f ( x i, y
j)
(1)
x 0 y 0
Where i = 0,1,,M-1
j = 0,1,,N-1 and the summation is taken over the
region in the image where w and f overlap.
(a)
(b)
IC
panel
Actual
reject
unit
Pattern
matching
Particle
Reject unit
in map file
Compared
IC position
result
80
80
80
Correct
17
17
17
Correct
11
11
11
Correct
14
12
14
Correct
42
41
42
Correct
(a)
(b)
(c)
(d)
(a)
(b)
Fig. 8 IC detect image (a) Pattern Matching technique detect
IC (b) Particle method detect IC.
IV. RESULT AND DISCUSSION
Table I shows the result of IC detection device with the
difference panel in case of actual IC number and position
matches with map file. Figure 9 shows the result from Table I.
The left hand sides of Fig. 9 are original image. The middle
images are graphic image which convert from number of map
file number 3 and 4 will be red dot and blue dot in map file
(e)
Fig. 9 Compare result in case actual IC image match with
map file.
Figure 10 shows the compare pictures of remaining IC
image and map file between actual IC number and position
mismatch with map file. The application is added red dot to
easier to see the mismatch position. The compared result
shows a capability in detecting IC device in difference panel
in case actual IC mismatch with map file shown in Table II.
TABLE II COMPARED RESULT IN CASE ACTUAL IC IMAGE MIS MATCH WITH MAP
FILE
IC
panel
Actual
reject
unit
Pattern
matching
Particle
Reject unit
in map file
Compared
IC position
result
10
Found
Mismatch
24
24
25
Found
Mismatch
Found
Mismatch
12
12
13
Found
Mismatch
10
Found
Mismatch
V. CONCLUSION
This paper presents an automatic method for compare the
object position in IC panel images. As shown in the result, the
proposed method successes to count and compare the object
positions with map file. The application accuracy is 100% in
the existing 10 IC panel images. However, this method also
can reduce human error and faster than human compare IC
unit positions
(f)
(g)
REFERENCES
[1]
(h)
[2]
[3]
[4]
(i)
[5]
[6]
(j)
Fig. 10 Comparison results in case actual IC image mismatch
with map file.
I.
INTRODUCTION
SYSTEM CHALLENGED
IV. SUMMARY
Fig. 1.
References
Date
18/1/2016
Activity
Lecture
Work shop
19/1/2016
20/1/2016
Lecture
Work shop
Assignment
21/1/2016
Assignment
22/1/2016
Assignment
Topics
MATLAB GUI & brief intro to
stat & machine learning toolbox
(distribution, cluster analysis)
STM32 programming and BLE
stack
local classroom computer
clicker device -> gateway +
screen <- smartphone
data synchronization
gateway database -> cloud server
analysis platform
MATLAB UI + cloud interface +
analysis
[1]
[2]
[3]
[4]
[5]
Y. Shi, W. Xie, G. Xu, R. Shi, E. Chen, Y. Mao, and F. Liu, The smart
classroom: merging technologies for seamless tele-education, IEEE
Pervasive Computing, Vol.2, pp. 47-55, April-June 2003.
C. Snow, J. M. Pullen, and P. McAndrews, Network EducationWare:
an open-source web-based system for synchronous distance education,
IEEE Transactions on Education, Vol. 48, pp. 705-712, Nov. 2005.
D. Pishva, Smart Classrooms Bring Top-Quality Education around the
Globe, in Proc. Int. Symp. On Applications and the Internet
Workshops, 2007, pp. 40-44, Jan 2007.
C. Di, Z. Gang, and X. Juhong, An Introduction to The Technology Of
Blending-reality Smart Classroom, in Proc. Int. Symp. On Knowledge
Acquisition and Modeling, 2008, pp. 516 519, 21-22 Dec. 2008.
G. G. D. Nishantha, D. Pishva, and Y. Hayashida, Smart Classrooms:
Architectural Requirements and Deployment Issues, in Proc. 2008
IEEE Region 10 and the Third int. Conf. on Industrial and Information
Systems, pp. 1-6, 8-10 Dec. 2008.
[6]
[7]
[8]
[9]
I. I NTRODUCTION
When talking about a cryptosystem, Alice, Bob, and Eve are
often mentioned as standard parties involved. In general, Alice
and Bob want to communicate with each other in secret, while
Eve wants to know what they are talking about. Traditionally,
Alice and Bob have no conflicting goal. They usually agree
on a specific encryption-decryption scheme. However, with
the advance in communication technology nowadays, such as
the Internet-of-Things (IoT) technology, Alice and Bob are
not just two computers anymore, they can be anything from
smartphones and tablets to wearable sensors or home washing
machines. Therefore, although Alice and Bob have the same
goal of communicating securely, they may be in conflict once
we talk about the data rate at which each party would like
to transmit and receive data, as well as the computational
complexity each party is willing to spend for encryptiondecryption processes.
If we make an assumption that each of Alice and Bob has
in his or her disposal more than one encryption/decryption
schemes, their choice of scheme can be adapted to the resources they have at the time secret messages are communicated, such as the transmission bandwidth and the CPU
resource.
To help illustrating the adaptive framework, we describe a
specific system that we have developed as a concrete example,
Application Server
ebook's
page no.
R : A1 A2 ... An ! ,
Hybrid AES-ECC
Encryption
ebooks storage
Web Service
Ebook Reader
Content Streaming
Smartphone
Fig. 1.
Laptop
{(HD , HC , HS ), (HD , HC , LS ),
(HD , LC , HS ), (HD , LC , LS ),
(LD , HC , HS ), (LD , HC , LS ),
(LD , LC , HS ), (LD , LC , LS )},
(1)
(3)
(4)
where !HD and !LD means high and low delay, !SC and
!F C means success communication and failed communication, respectively.
Formulation 2: Example of specification formulation at the
concrete level
RC : AC1 AC2 ! C ,
(5)
B. Secret Mixing
TABLE I
D IFFERENCES IN THE BASIC C OMPONENTS OF THE F OUR H YBRID
E NCRYPTION M ODELS U SED FOR A DAPTATION
Model
KEM
Symmetric Key
Encryption
KEM with
Secr. Mix. (B)
AES with
Secr. Mix. (D)
KEM with
Secr. Mix. (B)
AES with
Secr. Mix. (D)
(x, y) = RdA
(13)
(14)
dB
Zq ,
Zq ,
(7)
dA G,
(9)
QB
dB G,
(10)
1.3) Alice and Bob send her/his public key (Rs PK#1,
Ss PK#1) to each other as shown in Fig. 2. Only Alices
public key will be used in KEM. Bobs public key will be
used for creating the digital signature discussed in the system
implementation part.
2) Key Encapsulation of Bob:
2.1) Bob generates the shared secret x using a random
number r and Alices public key QA .
(x, y) = rQA = rGdA = RdA
dB,i
Zq ,
Zq ,
(15)
(16)
1.2) Based on each of dA,i and dB,i , Alice and Bob generate
her/his public key QA,i and QB,i , respectively, from the
generator G. Only Alices public keys will be used in KEM.
Bobs public keys will be used for creating the digital signature
discussed in the system implementation part.
(12)
(11)
Fig. 2.
(8)
QA,i
dA,i G,
(17)
QB,i
dB,i G,
(18)
1.3) Alice and Bob send her/his first public keys QA,1 (Rs
PK#11)and QB,1 (Ss PK#11) to each other as shown in Fig. 3.
(23)
Fig. 3.
(20)
(22)
(24)
3.3) With the mi derived from 3.2) and the known parameters aj,k ,j = 1, 2, ..., n, k = 1, 2, ..., n, the system of n
equations in (21) is solved to find KAES,i , i = 1, 2, ..., n.
3.4) All parts KAES,i , i = 1, 2, ..., n, are assembled into
the full AES key KAES .
As can be seen in Fig. 3, steps 1)-3) of KEM with secret
mixing are repeated periodically such that Alice can use the
AES Key Cipherblock # ji, j = 1, 2, ..., N , i = 1, 2, ..., n to
derive the ith partition of the j th updates of the symmetric
key KAES . The regular key updating leads to better security.
In our system, if the AES key is partitioned into four parts
before mixing, i.e., n = 4 in Eqs. (20) and (21). According to
[15], we can express the secret mixing in the following matrix
equation.
2
3 2
3 2
3
m1
1 1 1 1
KAES,1
6 m2 7 6 1 2 3 4 7 6 KAES,2 7
6
7 6
7 6
7
(25)
4 m3 5 = 4 1 3 6 8 5 4 KAES,3 5
m4
1 5 8 15
KAES,4
Note, however, that the proposed mixture is not the only
secret mixing possible. Any mixing, with the property that
the missing knowledge of any mixture left the enemy clueless
about every key partition, can be considered optimal.
C. Normal AES Symmetric Key Encryption
After Bob finishes key encapsulation in the KEM part, either
with or without secret mixing, he uses the encapsulated AES
key for symmetric key encryption of his plaintext.
D. AES Symmetric Key Encryption with Secret Mixing
The difference between normal AES encryption and AES
encryption with secret mixing can be illustrated in Fig. 4.
Unlike the normal AES encryption shown in Fig. 4(a), AES
encryption with secret mixing divides the secret message into
two parts, as shown in Fig. 4(b). Each symbol of the first
part, denoted by x1 , is encrypted by AES, whereas that of
the second part, denoted by x2 , is XORed with x1 bit by bit.
The encrypted version of x1 , represented by A(x1 ), and the
mixed information x1 x2 are transmitted to the receiver, who
derives x1 by decrypting the AES cipher and derives x2 by
subtracting x1 from the received x1 x2 .
Since this system encrypts only a half message with AES,
we can save the encryption and decryption time by about one
half of that in pure AES.
Note that, when compared with normal AES, the amount
of security decrease in AES with secret mixing is negligible.
Fig. 5. Test Result for the Case of 4 Key Partitions (n = 4) without Secret
Mixing (a) and with Secret Mixing (b)
Fig. 4. Comparison between (a) Normal AES Encryption and (b) AES
Encryption with Secret Mixing
Comparison
1. Security Level
2. Comput. Comp.
3. Data Rate
[8] R. Cramer and V. Shoup, A Practical Public Key Cryptosystem Provably Secure against Adaptive Chosen Ciphertext Attack, Advances in
Cryptology-CRYPTO98, pp. 13-25, Springer, 1998.
[9] R. Cramer and V. Shoup, Design and analysis of practical public-key
encryption schemes secure against adaptive chosen ciphertext attack,
SIAM Journal on Computing, vol. 33, no. 1, pp. 167-226, 2003.
[10] R. Ahlswede, N. Cai, S.-Y.R. Li, and R.W. Yeung, Network Information
Flow, IEEE Trans. Inform. Theory, vol. 46, no. 4, pp. 1204-1216, Jul.
2000.
[11] N. Cai and R.W. Yeung, Secure Network Coding, Int. Symp. Information Theory, Jun. 2002.
[12] K. Bhattad and K.R. Narayanan, Weakly Secure Network Coding, in
Proc. NETCOD, Apr. 2005.
[13] S.-H. Zhu, Research of Hybrid Cipher Algorithm Application to
Hydraulic Information Transmission 2011 Int. Conf. on Electronics,
Communications and Control (ICECC), Ningbo, Sep. 2011.
[14] W. Sriborrirux, A. Limmanee, and S. Kraipui, Unequal Security
Protection: A Unified Framework, Implementation, and Performance
Evaluation of Theoretical and Practical Security Int. Symp. on Wireless
Communication Systems (ISWCS), Ilmenau, Aug. 2013.
[15] A. Limmanee, W. Sriborrirux, and S. Kraipui, Hybrid Encryption
Scheme for Digital Content with Key Partitioning and Secret Mixing:
Design and Implementation Int. Symp. on Wireless Communication
Systems (ISWCS), Ilmenau, Aug. 2013.
Nakrom Massaya
Jirayut Phangklin
I. I NTRODUCTION
Health monitoring is one of smart services in smart cities
that integrate multiple information and communication technology interacting with people to enhance quality of life.
Health monitoring application is not only for structural health
monitoring [1] but also for biomedical engineering [2]. Health
monitoring can be a key performance indicator reflecting a
cooperative between a smart city and quality of life. The
possibility of two events in a smart city can be as follows.
In structural health monitoring, when vibration of a bridge
surpasses a threshold, alarm messages will be reported to
emergency department via wireless sensor networks [3], [4]
and [5]. In healthcare monitoring, while people are exercising,
their heartbeats are sent to a healthcare centre via connected
devices. Based on two such events of health monitoring systems, sensor circuit and signal analysis are important because
they can improve accuracy of measurement. In order to narrow
down the list of possible events, let this paper focus on
heartbeat measurement.
Heartbeat measurement is a contemporary topic of health
monitoring research since it directly affects quality of life.
Heartbeat can be measured by counting the number of contractions of the heart per minute. This depends on the bodys
physical needs, including the need to absorb oxygen and excrete carbon dioxide. It is normally close to the pulse measured
at any peripheral point. Heartbeat is obviously relevant to
everyday activities such as physical exercise, sleep, anxiety,
stress, illness, and ingestion of drugs. This motivates many
researchers to study how it works. Even though existing work
proposed a variety of heartbeat measurement techniques but
they incompletely explain in terms of analog and discrete
signal analysis.
Fig. 1.
Fig. 2. Input and output of heartbeat signals reshaped by (a) the 1st state
of the HPF, (b) the 1st state of the BPF, (c) the 2nd state of the HPF and (d)
the 2nd state of the BPF (heartbeat pulses)
1
2R3 C1
(1)
fc,LP F 1 =
1
2R4 C2
(2)
R4
R5
(3)
Fig. 3.
Discrete-time signals
n
X
yi
(4)
i=1
m
X
xj
(5)
j=1
hL+1
...
hM
A trajectory matrix H = [H1 ,H2 ,. . . ,HK ] can be decomposed into the multi-dimensional series H1 , H2 , . . . , HM with
vector Hi = (hi , . . . , hi+L 1 ) where K = M
L + 1.
Alternatively, H can be written in (8).
H=
d
X
Hi ,
d = min {L, K}
i=1
(8)
g
X
HIp
(9)
p=1
1X
hi
t i=1
(10)
HR =
60, 000
i
BPM
(11)
TABLE I
P ERFORMANCE OF HEARTBEAT MEASUREMENT
Index
i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
IBI
i
(ms)
750
738
768
732
740
722
714
724
724
760
706
722
722
750
732
740
700
700
750
750
732
698
706
714
698
Measured
HRo
(BPM)
80
81
78
82
81
83
84
83
83
79
85
83
83
80
82
81
78
78
80
80
82
86
85
84
86
Standard
HRs
(BPM)
80
80
81
82
82
81
82
84
83
82
82
82
81
82
83
80
79
79
80
82
83
85
86
85
85
Absolute error
|HRs HRo |
(BPM)
0
1
3
0
1
2
2
1
0
3
3
1
2
2
1
1
1
1
0
2
1
1
1
1
1
B. Experiment results
During experiment, when a finger is placed on the heartbeat
sensor, a first valid heartbeat can be detected after 5-10 seconds. Then their readings are periodically changed according
to human heart rates. In Fig. 4, heart rates of our implemented
circuit HRo are compared to that of a standard instrument
HRs [12]. A round of heartbeat measurement consists of W =
25 readings. In a first round, 25 readings are shown in Table I
where a mean of absolute errors is 1.28 BPM and standard
deviation, , is 0.89. IBIs are between 698 to 768 milliseconds
(1.43 - 1.30 Hertz). A time series of absolute errors H can be
written as equation (12).
E=|
W
1 X
HRs
W i=1
W
1 X
HRo |
W i=1
Fig. 5.
of heartbeat measurement
(12)
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
health monitoring, IEEE Sensors Journal, vol. 15, no. 2, pp. 809818,
Feb 2015.
Z. Zhang, Z. Pi, and B. Liu, Troika: A general framework for
heart rate monitoring using wrist-type photoplethysmographic signals
during intensive physical exercise, IEEE Transactions on Biomedical
Engineering, vol. 62, no. 2, pp. 522531, Feb 2015.
A. Boonsongsrikul, S. Kocijancic, and S. Suppharangsan, Effective energy consumption on wireless sensor networks: Survey and challenges,
in Information Communication Technology Electronics Microelectronics
(MIPRO), 2013 36th International Convention on, May 2013, pp. 469
473.
A. Boonsongsrikul, S. K. Park, and S. H. Shin, An adaptively balancing
workload protocol on query trees for maximizing lifetime in sensor
networks, in Advanced Communication Technology (ICACT), 2012 14th
International Conference on, Feb 2012, pp. 195200.
A. Boonsongsrikul, K. suk Lhee, and M. Hong, Securing data aggregation against false data injection in wireless sensor networks, in
Advanced Communication Technology (ICACT), 2010 The 12th International Conference on, vol. 1, Feb 2010, pp. 2934.
H. Mansor, S. S. Meskam, N. S. Zamery, N. Q. A. M. Rusli, and
R. Akmeliawati, Portable heart rate measurement for remote health
monitoring system, in Control Conference (ASCC), 2015 10th Asian,
May 2015, pp. 15.
S. R. Mohana and H. V. R. Aradhya, Remote monitoring of heart
rate and music to tune the heart rate, in Communication Technologies
(GCCT), 2015 Global Conference on, April 2015, pp. 678681.
M. T. I. Papon, I. Ahmad, N. Saquib, and A. Rahman, Non-invasive
heart rate measuring smartphone applications using on-board cameras:
A short survey, in Networking Systems and Security (NSysS), 2015
International Conference on, Jan 2015, pp. 16.
M. Ghamari, C. Aguilar, C. Soltanpur, and H. Nazeran, Rapid prototyping of a smart device-based wireless reflectance photoplethysmograph,
in 2016 32nd Southern Biomedical Engineering Conference (SBEC),
March 2016, pp. 175176.
M. A. Miah, M. H. Kabir, M. S. R. Tanveer, and M. A. H. Akhand,
Continuous heart rate and body temperature monitoring system using
arduino uno and android device, in Electrical Information and Communication Technology (EICT), 2015 2nd International Conference on,
Dec 2015, pp. 183188.
G. Pang and C. Ma, A neo-reflective wrist pulse oximeter, IEEE
Access, vol. 2, pp. 15621567, 2014.
S. Smiths Medical PM. (2008, May) Bci3301hand-held pulse oximeter.
[Online]. Available: https://goo.gl/25yKdn
I. INTRODUCTION
The term zero energy buildings, recognized as net zero
energy buildings are about the buildings thathave zero carbon
emissions on annual basis. The daily energy requirements and
generate energymust be minimized by using renewable
resources [1].
The concept of zero energy buildings is completely about
minimizing the required energy use, particularly energy that
generated from fossil fuel and implementing the use of
renewable energy resource to meet the energy demand.
supported by Danny H.W. Li et al [2].
Energy consumption can be minimized by applying new
technologies in the buildings such as wall materials, insulation
materials, high efficient air conditioning system or solar
lighting which lead to meet zero energy building designs [3].
Farajallah Alrashed [4] mentioned that zero energy buildings
actually rely on renewable energy to meet their energy needs,
therefore, it isimportant to have sufficient and appropriate
renewable resources to sustain the energy demands and
photovoltaic or PV seems to be one of the most sustainable
energy generation technologies for zero energy building [5].
Solar Cell
Controller
Battery
Inverter
Energy
Usage
Cabinet
8
7
6
5
4
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
AVERAGE SOLAR
RADIATION (KW/M2-DAY)
MONTHS
Equipment
No. of
equipment
Power of
Equipment (W)
Daily
use (h)
Light 1
Light 2
Light 3
Fan
1
1
1
1
12
12
12
58
1
1
12
18
Computer
50
Total
Energy
required
(kWh/d)
0.012
0.012
0.144
1.044
0.05
1.262
Controll
er
DC-DC
2,411
Wh
Inverter
DC-AC
1,325
Wh
Load
1,262
Wh
1,086
Wh
Battery
2,880
Wh
Loss
127
Wh
(a)
Loss
66 Wh
Calculation.
A. Equations
As the concept of off-grid PV system, the electricity that
generated from PV power generation must be sufficient to
support the energy demand and must be able to preserve in the
battery in order to use the electricity when the PV system
cannot generate the electricity because of no sunshine. The
equation for the generated energyis shown in (1)
(b)
(1)
(3)
(4)
Noted that, losses in the system such as Ohmic loss has not
been calculated yet.
B. Scenario 1: full-day sunshine
Controlle
r
DC-DC
1,208
Wh
Loss
64 Wh
117
Wh
Battery
2,880
Wh
(a)
Calculation.
Inverter
DC-AC
1,325
Wh
Loss
66 Wh
Load
1,262
Wh
(b)
D. Scenario 3: no sunshine
In the raining day that PV array cannot generate electricity,
it is necessary to use the energy that saved in battery. If the
battery is full charged, it is possible to use the energy for 1.7
days as shown in Figure 6aand the forecasted load profile is
shown in Figure 6b.
Inverter
DC-AC
1,325
Wh
Controll
er
DC-DC
0 Wh
Solar
Cell
0 Wh
Load
1,262
Wh
ACKNOWLDGMENT
Authors would like to thank Dr. Bhirasak Vorasutharosoth
to permit and support the installation of PV system, battery,
inverter and data analysis system to collect useful data in his
meditation home.
REFERENCES
[1]
1,325
Wh
Battery
2,880
Wh
Loss
0 Wh
(a)
[2]
Loss
66 Wh
Calculation.
[3]
[4]
[5]
(b)
(c)
Fig. 6. No sunlight.
[6]
E. Energy ratio
Full-day sunlight is the best case that used to find the
energy ratio. This system can generate electricity 2,538 Wh
with 1,262 Wh load requirement and the remaining 1,086 Wh
is saved in the battery. Without concerning other losses in the
system, such as Ohmic loss, Energygen : Energyload :
Energystoragewill become 2:1:1.
V.
CONCLUSION
[7]
[8]
[9]
I. INTRODUCTION
At present, automation system is popular in the industry
due to the effect of competition in the market. An important
factor is to deliver the products in time with no man day labor
cost. AGV safely transport is operated without human
intervention within production, logistic, warehouse, the clear
way to reduce costs and to increase efficiency and
profitability. If considering the cost of process production is
dominated by materials handling. Therefore, we can use the
AGV to work in the industry. It can increase production
capacity on process, reduce labor cost and also provide
flexibility in working.
The AGV is designed as a prototype using brushless DC
motor which has high efficiency, smaller than induction
motor, light weight and less maintenance. In this paper the
objectives are to improve precision of navigation system, to
improve AGV control, monitoring system and to improve
reliability of AGV. For the scope of work using gear wheel
hub brushless DC motor 750W rated power, 100-700 rpm 36
Vdc. The AGV is functioned and control by microcontroller
ATMega2560, ATTiny85, Nano328 and dsPIC30F2010 using
a. CPU
performance
emergency stop
45, 90 corner
30 cm ultrasonic stop
10 cm round curve
siren mode
warning sound
80 kg payload
battery status
simplicity of charging
home/end setup
IV. CONCLUSION
The AGV are designed to meet the safety and performance
of industry requirement. Aluminum tape is a solution of
tracking guide path with low cost. The AGV system designed
is proposed. The prototype need to be improved on the
material as comment from the industry feedback
REFERENCES
a. AGV sensors
Figure 3 BLDC motor drive
[1]
[2]
CCS Concepts
CCS Hardware Communication hardware, interfaces and
storage Wireless integrated network sensors.
Keywords
Industrial Internet of Things; Wireless Control; Arduino
Microcontroller; MQTT Internet Protocol; Android.
1. INRODUCTION
Digital transformations have been of much interested by industries
and enterprises as an emergence of intelligent devices, ubiquitous
internet connectivity, big data analytics and cloud computing has
potentially driven new business opportunities, reduced waste and
costs, and eliminated loss regarding human-machine (H2M) or
machine-machine (M2M) activities. As a result from the increase
in digital transformations, the Internet of Things (IoT) has
consequently played an important role as a new segment in
existing platforms. Such an IoT can generally be defined as a
system in which the internet, communicating through Internet
Protocol (IP), is connected to the real world via ubiquitous sensors
and actuators, integrating diverse data sets obtained from physical
sensors and Information Technology (IT) infrastructures in order
to enable intelligent data analytics and control processes.
Recently, the IoT has extensively applied to various industries
such as healthcare [1], transportation [2], agriculture and water
supply [3-4], smart home and city [5-6], and manufacturing [7].
Cloud Server
MQTT
Router
ESP8266 Wi-Fi
Module
MQTT
Router
Router
Web-Based
Application on PC
Android-Based
Application on
Smart Phone or
Tablet
Arduino
Compatible Board
(a)
Industrial
Input / Output
via PLC
Built-In RS232C
Built-In USB
Modbus RTU
Pulse Outputs
Inverter Inverter
(b)
Figure 3. Proposed Arduino compatable board; (a) Circuit
diagram of a Arduino ATMEGA328P, (b) an implemented
board with full periferals for Industrial Input/Ootput ports.
PLC Module
Power Supply
Wi-Fi
Neumatic Switches
(a)
(b)
(c)
(d)
ACKNOWLEDGEMENTS
The data involving the control signals and the monitroing states
are sent to the router via MQTT protocal. The data is connected to
the samrt devices such as tablets or smart phones through the
integrated Wi-Fi module. Figure 5 shows the overall system
design, including a PLC, an Arduino Compatible board, a Wi-Fi
module, a power supply, and a neumatic swutches. As aresult of
Android designs, Figure 6 shows the display on the Android
Application, (a) a sleep mode (No machine is ON), (b) all
machines is ON, (c) a virual operating machine of 20 inputs, and
(d) the setting functions. It was apparent that this application
could control all devices very accurately.
REFERENCES
[1]
CONCLUSION
A raodmap towards industrial 4.0 in Thailand and ASEAN
countries has aroused the need for Industrial Internet of Things
that enables remote communications via wireless IP devices and
interactions between human and manufacturing processes for the
enhancement of performances and efficiency as well as the
reduction in manufacturing costs. This paper has presented a
complete IIoT platform in which the universal PLC is controlled
by an open-software Arduino microcontroller via a high-voltage
peripheral board. The communication between PLC and end users
was achieved by Wi-Fi technology and MQTT internet protocol
that allow data storage and monitoring in cloud server. The GUI
was designed for both a web-based display and an Android
application. Demonstrations on the pneumatic control system
were also included. The proposed system offers a low-cost highefficiency implementation of IIoT for Small-and-Medium
Enterprises (SMEs).
Suranan
Noimanee
and
Somkiat
Wattanasirichaigoon,
Implementation of Vital Signs Monitoring System Using Wireless
Networks, International Journal of Applied Biomedical
Engineering, Vol.1, No.1, 2008.
**
International College, King Mongkuts Institute of Technology Ladkrabang, Bangkok 10520, Thailand
Department of Biomedical Engineering, Srinakharinwirot University, Ongkharak, Nakhon Nayok 12110, Thailand
***
Department of Electrical Engineering, Mahasarakham University, Mahasarakham 44150, Thailand
E-mail: montri.ph@kmitl.ac.th, suchadat@g.swu.ac.th, apinan.a@msu.ac.th
I. INTRODUCTION
Watermelon (Citrullus lanatus) is commonly grown and
consumed in all tropical areas of agricultural countries. For
example, in Japan, watermelons are popular and also have a
shape to provide value-added. In China, watermelons are the
top producer in the world. In Thailand, the country is in tropical weather which suitable for planting rice, fruit and vegetable, also the one of major food producer of the world each
year. Watermelon is also a tropical economic cucurbit crops
grown throughout the world. One of the main factors in ensuring the consistent marketing of watermelon is the quality of
the product. Undoubtedly, one of the most common problems
is that determining watermelon ripeness based on visual inspection. Traditional methods for assessing fruit ripeness are
destructive, thus cannot be so readily applied, particularly in
mass production [1].
For the related works, there were many non-destructive
methods proposed to determine the ripeness of fruits, e.g.,
durian [2], mango [3], potato [4], apple [5], orange [6] and
watermelon [7]. Therefore, it is important to monitor and
control the fruit ripeness since it is a major issue in fruit production.
Watermelon
100 cm
Digital camera
Input image
(*.BMP)
Fig. 1. Watermelons rind image acquisition in the experiment.
Input image
Normalization
Binary
Transformation
Morphological
Operation
1) Normalization: color image will be transformed into grayscale image. Due to the diversity of images in terms of contrasting or strengthening during acquisition process, the normalization is proposed to overcome that changes the range of
pixel intensity values to bring the image into a same range of
expected values, i.e., 0 255 in 8-bit resolution of this experiment. Normalization transforms an n-dimensional grayscale
image (n = 8) into the following equations:
Region of Interest
Classification
Feature Extraction
(
RGB Color
Segmentation
Rind Texture
Analysis
Background
)
(
(
)
(
(
)
)
(1)
(2)
where
(
) and
(
) are the minimal contrast function and maximal contrast function, respectively.
Output of this process is the image shading correction.
2) Binary Transformation: after the normalization process,
image will be transformed to a binary image based on the
histogram value thresholding. The histogram of image is
initially segmented into two parts using an average value of
(
) where
contrast value in each pixel, i.e.,
is a normalized contrast value ranged 0 1. Typical example of histogram based thresholding as shown in Fig. 3. The
binary transformation can be determined on the basis of the
following rule:
Object (watermelon)
T = 0.78
Fig. 3. The histogram of the grayscale image and the corresponding threshold
value to separate watermelon from its background.
(3)
(4)
where
and are the structuring element of erosion and dilation, respectively. In our experiment, we set erosion
of 55
all-ones matrix and dilation
of 99 all-ones matrix.
4) Region of Interest Classification: this is called ROI classification, the output image can be obtained by the multiplication of image from opening operation. Therefore, the ROI of
image is a watermelon extracted from the background. For
examples, all results passed the proposed method can be seen
in Table I.
Result
Input Image
Normalization
Fig. 4. Expected values shown in vertical dotted line in RGB space of color
segmentation using k-means clustering where k is 3.
Color Cluster
Morphological Operation
G1 Dataset (Ripe)
Pixel
[R, G, B]
42.33 %
33.44 %
24.23 %
G2 Dataset (Unripe)
ROI Classification
48.51 %
27.13 %
24.36 %
(* p < 0.001)
Fig. 7. Feature extraction of the standard deviation (S.D.) and its moving
average was applied to determine the ripe and unripe data. The one-way
ANOVA was applied to test statistically significant among the groups.
Fig. 5. Experimental results of rind texture analysis by using LoG based edge
detection corresponding to ripe watermelons.
6) Rind Texture Analysis: texture is characterized as the spatial distribution of the gray degree in a set of adjacent pixels.
Therefore, in this paper, we also investigate the rind texture of
watermelon since it considered as a unity depending on the
type of varieties. An image classification process is generally
requires the stage of feature extraction, which we proposed
the edge detection of the grayscale ROI image using zerocrossings after filtering with a Laplacian of Gaussian (LoG)
filter, i.e., MATLAB function2 edge(img,'log') in the Image
Processing Toolbox. The obtained results of rind texture analysis corresponding to ripe and unripe watermelons shown as
Figs. 5 and 6, respectively.
III. EXPERIMENTS AND RESULTS
Fig. 6. Experimental results of rind texture analysis by using LoG based edge
detection corresponding to unripe watermelons.
1)
2)
REFERENCES
[1] E. Llobet, E.L. Hines, J.W. Gardner and S. Franco, Nondestructive
banana ripeness determination using a neural network based electronic
nose, Meas. Sci. Technol. vol.10, no.5, pp.538-548. 1999.
[2] M. Phothisonothai, Nondestructive maturity classification of durian
based on fractal features, Int. Symp. Signal Processing and Its Applications (ISSPA), pp.566569, 2010.
[3] T. Charoenpong, K. Chamnongthai, P. Kamhom, and M. Krairiksh
Volume measurement of mango by using 2D ellipse model, IEEE Int.
Conf. Industrial Technology (ICIT), vol.3, pp.14381441, 2004.
[4] M. Barnes, T. Duckett, and G. Cielniak, Boosting minimalist classifiers
for blemish detection in potatoes, Int. Conf. Image and Vision Computing New Zealand, pp.397402, 2009.
[5] J. Zhou, H. Yin, J. Liu, and L. Fan, Method of image fusion for apple
surface quality detection, Automatic Control and Artificial Intelligence
(ACAI 2012), pp.13391342, 2012.
[6] C.P.D. Carolina and N.T.D. David, Classification of oranges by maturity, using image processing techniques, International Congress of
Engineering Mechatronics and Automation (CIIMA), pp.1-5, 2014.
[7] F. Y. A. Rahman and et al, Monitoring of watermelon ripeness based
on fuzzy logic, WRI World Congress on Computer Science and Information Engineering (CSIE '09), vol.06, pp.67-70, 2009.
[8] M.N. Salim and Murinto, Image processing application for detecting
the ripeness of watermelon based on features of the rind texture, Int.
Conf. Green World in Business and Technology (ICGWBT), 2014.
[9] C. Connolly and T. Fleiss, A study of efficiency and accuracy in the
transformation from RGB to CIELAB color space, IEEE Trans. Image
Processing, vol.6, pp.10461048, July 1997.
[10] Watermelon ripeness selection, wikiHow, http://th.wikihow.com (in
Thai)
ACKNOWLEDGEMENTS
This research was partially supported by the International
College (IC), King Mongkuts Institute of Technology Ladkrabang (KMITL). The authors would like to thank students
from Department of Electrical Engineering, Mahasarakham
University, who collected all images in the experiment.
The authors would like to thank the academic collaboration project between
the Electricite du Laos (EDL) and the Electricity Generating Authority of
Thailand (EGAT) for financial support for this research.
Khambai Vongsomphone (e-mail: khambai_vsp@hotmail.com) and
I. INTRODUCTION
C1 C2 ... Cn
(1)
Ci ( PGi )
i 1
PGT
PG1
(2)
PGi
i 1
n
PGi
PD
(3)
i 1
MinPGi
Ci ( PGi )
MinPGi
i 1
PGi
PGi2 )
(4)
i 1
where:
n : The number of generator units,
i : Index of dispatch able units,
Ci : The cost generation of units i, [$/h]
CT : The total cost generation in system, [$/h]
PGi : The power generation of unit i, [MW]
PGT : The total power generation in system [MW]
i, i, i : Coefficients for power generation cost of unit i
A. Power balance
The optimal dispatch power generation of hydro power
plants should be such that total electric power generation equal
to the total of power transmission line loss plus total system
load demand.
That is:
n
(5)
PG P P
i
i 1
Input
(X)
Neural
Network
(H)
Start
Set lambda
Adjust weights
Calculate PGi
For i= 1n
Compare
First iteration
Target
Calculate error
=Pd - PGi
Adjust Lambda
Output
(Y)
Yes
No
No
Absolute error
tolerance
Yes
Print out lambda, power
generation & total cost
End
Fig. 3. Step of optimal dispatch by the lambda-iteration method.
Hj
( X i wij ), j= 1, 2, . . . n
(7)
i 1
A. Backpropagation algorithm
Yk
( H j v jk ), k= 1, 2, . . . n
(8)
j 1
PG1
V1-1
h1
W1
PG2
Input load
h2
PLoad
PG3
W20
PG4
V20-5
h20
Input layer
Hidden layer
PG5
Output layer
i PGi
(9)
PGi2
SIMULATION RESULTS
($/h)
($/MWh) ($/MW2h)
No.
Load
(MW)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
Limit of generation
Min
(MW)
Max
(MW)
NNG1 (PG1)
442.39
54.109
0.0087
15.00
155.00
NL (PG2)
642.57
54.381
0.0200
20.00
60.00
NM 3 (PG3)
775.62
52.609
0.0198
10.00
40.00
NL 1/2 (PG4)
329.37
48.996
0.0154
20.00
100.00
NNG5 (PG5)
348.51
47.635
0.0204
20.00
120.00
NL
NM3
NL
NNG5
Cost
($/h)
(PG1)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
20.00
40.00
60.00
60.00
74.21
88.20
102.19
116.17
130.16
144.14
160.00
(PG2)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
20.00
25.79
31.80
37.82
43.83
49.84
55.86
60.00
(PG3)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
20.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
(PG4)
39.05
50.42
61.79
73.17
84.54
95.91
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
(PG5)
60.95
69.58
78.21
86.84
95.46
104.09
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
7,457
8,465
9,479
10,500
11,529
12,564
13,609
14,669
15,746
16,831
17,922
19,019
20,120
21,225
22,336
23,452
24,572
25,697
26,827
27,962
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Load
(MW)
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
NL
(PGn2)
NM3
(PGn3)
NL
(PGn4)
NNG5
(PGn5)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
20.00
40.00
60.00
60.00
74.21
88.20
102.19
116.17
130.16
144.14
160.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
20.00
25.79
31.80
37.82
43.83
49.84
55.86
60.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
20.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
40.00
39.05
50.42
61.79
73.17
84.54
95.91
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
60.95
69.58
78.21
86.84
95.46
104.09
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
Cost
($/h)
7,457
8,465
9,479
10,500
11,529
12,564
13,609
14,669
15,746
16,831
17,922
19,019
20,120
21,225
22,336
23,452
24,572
25,697
26,827
27,962
Total load
V. CONCLUSION
The This paper has presented the optimal dispatch of hydro
power plant scheduling. It provides the technical method to
optimally determine generation in the central-1 of EDL of Laos.
The objective function is to minimize of generation costs. This
paper proposes the neural network and lambda iteration
method, which is used for finding the optimal dispatch of the
hydro power generation. The simulation results of both methods
are very similar. However, the neural network method uses less
time than the lambda iteration method. When compared
between actual operation and the technical method. We can
observe that the optimize operation cost is lower than actually
operation cost.
ACKNOWLEDGMENT
PG1, PGn1
PG5, PGn5
PG3, PGn3
PG4, PGn4
PG2, PGn2
REFERENCES
[1]
[2]
[3]
Fig. 6. The result of optimal dispatch by using Lambda-iteration and Backpropagation neural network method.
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
Reliability
assessment;
reliability
improvement; investment.
I.
INTRODUCTION
Month
1
0
0
0
0
1
1
0
1
1
0
0
2
0
1
0
0
1
2
0 0
5 9
1 0
0 1
1 1
7 11
9 10 11 12
0 3
6 2
0 0
1 4
2 4
9 13
0
1
1
7
7
16
0
0
0
1
0
1
0
1
0
0
0
1
0
1
0
1
1
3
0
0
0
0
0
0
Total
Interruption
outage
time (Hrs/yr)
(No/yr)
3
0:42
27
12:54
3
2:29
15
1:51
18
4:53
66
22:49
TABLE II
STATISTIC FAILURE RATE OF THE EACH FEEDER IN THE KHUATINUENG
SUBSTATION
Feeder
Failure Frequency
(1/a. km)
Number of outage
(Time)
Outage
duration
(min)
F1
F2
F3
F4
F5
0.90
2.76
0.84
0.66
7.62
3
27
3
15
18
25.2
752.4
137.4
90.6
371.8
RELIABILITY IMPROVEMENT
Reliability improvement
Y
Summerize
III.
STUDY SYSTEM
Nam Ngum 1 HP
F3. 5
2x17.5 MW
4x45 MW
Y
Khuatinueng SS
Y
115/22 kV
Y
2x12.5 MW
F2. 3
Y
Y
F3. 4
F2
F3. 2
F3
F4
F5. 1
F5
To Xiengnguen District
F2. 2
F1. 1
To Chomphet District
To Khai District
F5. 2
F3. 1
F1. 2
F2. 1
Phonexai District
F3. 3
F1. 3
F1
F5. 3
Pak-ou District
To Viengkham District
F5. 4
Pakseng District
F4. 1
II.
Breaker
LBS existing (Close)
LBS existing (Open)
LBS is a new installation (Close)
LBS is a new installation (Open)
Fig. 2. Single Line Diagram of New LBS Installation Equipment and LBS
Existing of the Electrical Distribution system in Khuatinueng
Substation [1].
Load (MW)
Number of Customer
F1
F2
F3
F4
F5
4.81
3.15
3.51
1.56
0.42
4,930
11,324
6,078
2,253
6,647
15
199.5
14
11
153.6
Traffic
accident
11%
Animal
9%
natural
disasters
29%
equipmen
t failure
17%
Feeder
F1
F2
F3
F4
F5
Tree
34%
Customers
serviced
(Unit)
4,423
8,171
5,879
2,038
4,005
Customers
interruptions
(Unit)
11,044
2,319
38,970
5,958
2,020
IV.
Failure
rate
(1/a.km)
0.6
0.7
0.6
0.6
0.8
Outage
duration
(Hours/yr)
0.42
12.54
2.29
1.51
4.23
CASE STUDY
Case 2
Case 3
Case 4
Case 5
0.1580
V.
0.316
2.00
30,753.92
SAIFI
(No/year)
SAIDI
(Hr/year)
CAIDI
(Hr/No)
Outage cost
($/year)
Case 1
Case 2
Case 3
Case 4
Case 5
0.0037
0.0744
0.0048
0.0072
0.0872
0.007
0.135
0.010
0.011
0.165
1.820
1.640
1.594
1.532
1.890
79,235.45
27,331.78
59,435.54
43,344.28
24,510.81
136,561.60
33,331.44
27,331.78
5.55
Case 3
227,602.67
74,574.08
59,435.54
4.92
Case 4
45,520.53
52,858.88
43,344.28
4.78
Case 5
182,082.13
30,753.92
24,510.81
29.16
CAIDI =
Outage cos t =
where
i i
[2]
[3]
[4]
[5]
[6]
[7]
(1)
l N r
i ii
NT
IER Pj
(4)
tj
IER =
NT
(3)
i i
REFERENCE
[1]
APPENDIX
l N
i ii
CONCLUSIONS
l N r
l N
SAIFI =
(2)
Daranee Hormdee*
I. INTRODUCTION
Due to its inexpensive cost, nowadays, a TFT LCD screen has
become a default device for user interfacing in an embedded system.
Supporting both text and colored graphic makes LCD display so user
friendly.
Displaying on LCD screen has been controlled by a microcontroller
system, so microcontroller must save all images in its within its internal
memory. Nonetheless, low level microcontrollers usually have low
capacity of memory, which might not be enough for saving all images
in case of large files. Therefore, the concept of low-complexity lossless
image compression was created to reduce the data size, and it should be
able to make image reconstruction with low-performance
microcontroller.
Data compression is very significant to multimedia data
compression. Currently, there are several methods of data compression
both in lossless and lossy. Lossless data compression such as Huffman
Coding, LWZ or Deflate, which can be used to measure frequency of
each alphabet appearing and replace with shorter data or replace phrase
appearing the second time with phrase appearing the first time in order
to reduce data size. These methods have been used with ZIP programs
for reducing the size of documents including lossless image in any
formats such as GIF [1] or PNG [2]. The downside of lossless data
compression is that a lot of memory is required due to dictionary table
created during decoding.
On the other hand, Lossy image compression is used to compress
image such as Joint Photographic Experts Group (JPEG, JPG) via
Discrete Cosine Transform (DCT) technique [3]. This DCT is be used
to convert the signal (spatial information) into numeric data (frequency
domain) so that the images information exists in a quantitative form
that can be manipulated for compression. However, inverse DCT
process is rather complicated to deal with.
From those reasons explained above, both traditional lossy and
lossless data compressions seems to be too complex on a low
performance microcontroller system.
This paper suggests a novel method of lossless image compression
which can be decoded on a conventional low performance embedded
system. The concept of this work is based on the characteristic of
images; clustering pixels often be same color, then reducing data
redundancy by grouping these same color pixels into objects;
Rectangle, Line and Point.
*Corresponding Author
ICESIT2016 Page 109 of 119
C. Related Works
Three previous relevant research works are listed in this section.
First research was about an image compression based on regionsegment technique which also used in this paper. The other 2 studies
focused on an image compression for low capacity and complexity
devices using run-length encoding technique.
Li Zhe-lin, et al suggested the new method of reducing the size of
cartoon image and published in 2009 [9]. Each color in an original
image was classified into a region. By deleting data inside except the
contour, could save data space in each region as default value: RGB-XY with Freeman Chain Code, and then compressing the acquired data
with Lempel-Ziv, further yields to the smaller size of data.
Roman Slaby, et al suggested the image compression as a
monochrome for displaying on LCD with low capacity devices and
published in 2013 [10]. Run-length Encoding technique was used for
white and black pixels in order to decrease data complexity.
In 2014, Bojan Stankic, et al suggested new file format used for
images named Embedded Run-length Image Encoding (ERLE) [11].
This method has high compression rate with low complexity of
Row_Index = 0
Start_Col_Index = 5
End_Col_Index = 6
Rect_Height = 7
Row_Index = 0
Start_Col_Index = 2
End_Col_Index = 3
6
(a)
0
(b)
3
6
(c)
(d)
Fig. 1. (a) Original Image, (b) Region and Objects divided for Pink, (c) Yellow
and (d) Brown.
Start Col
End Col
Start Col
End Col
Height
Color
TABLE I.
Start
2
4
1
0
2
4
1
4
3
Pink
(#FF80
80)
Yellow
(#FFFF
00)
Brown
(#8050
00)
Write Row_index
0
1
4
0
3
-
4
-
1+1
1+1
5
9
2+1
0
4
0
4
4
4
4+4
1
-
14+2+1
Repeat for the next Row_index until complete for all rows.
Repeat for the next color until complete for all colors
Selecting Row-Region process can begin, which only RowRegion containing less than 4 pixels will be rejected. The selected rows
was shown in the last column of Table I. Then, all selected rows are
saved in the data structure shown in Fig. 3.
The data structure of the proposed method starts with Image
Header and followed by Color-Region of the image (Fig. 3 (a)). Image
Header consists of Image_Width, Image_Height and Data_Length of
Color-Region (Fig. 3 (b)).
Color-Region is the region of each color within the image (Fig.
3 (c)). Color value is saved in RGB format, followed by Row-Region
data and end with Ending code of color, 0ffh (Fig. 3 (d)). Row-Region
data is listed with selected Row-Regions of the color in ascending order
of Row Index (Fig. 3 (e)).
Object is a group of clustering pixels found in each RowRegion, which can be divided into 3 types: Point, Line and Rectangle
(Fig. 3 (f)). All objects are saved to the list according to the type in
ascending order of Column Index, and following by Ending code of
object, 0ffh (Fig. 3 (g, i, k)).
For Point, Column_Index saved to list (Fig. 3 (l)).
For Line, Start_Column_Index and End_Column_Index are saved
to the list (Fig. 3 (j)).
End
TABLE II.
Condition
0
<= value < 128
128 <= value < 16,384
value >= 16,384
Data format
1 Byte Format, with setting MSB as 0
2 Bytes Format, with setting MSB as 1
Not supported
TABLE III.
Color-Region Data
(Step 1-2)
Image Header
Width
Height
Color #1
Color #...
Row Index
Color #...
Color #n
Row-Region #1
Rect #...
Start_Column_Index
Rectangles
(f)
End of Rects
(g)
Rect_Height
(h)
End of Lines
(i)
End_Column_Index
(j)
Rect #n
End_Column_Index
Line #1
Line #...
Line #n
Start_Column_Index
Point #...
(d)
(e)
Lines
Rect #1
Point #1
End of Colors
(c)
Row-Region #n
Row-Region #...
Points
(a)
(b)
Color #1
G
End of Points
(k)
Column_Index
(l)
Point #n
0
0
1
2
Data Structure
Compressed Image
Image Header
Color #1 (Pink)
Row #0
List of Point
End of Point
List of Line
End of Line
List of Rectangle
End of Rectangle
End of Color
Color #2 (Yellow)
Row #2
List of Point
End of Point
List of Line
End of Line
List of Rectangle
End of Rectangle
Row #4
List of Point
End of Point
List of Line
End of Line
List of Rectangle
End of Rectangle
End of Color
Color #3 (Brown)
Row #3
List of Point
End of Point
List of Line
End of Line
List of Rectangle
End of Rectangle
End of Color
TABLE IV.
Fig. 4. Original Image after all pixels in selected rows have been marked X
Location
(Row, Column)
(0, 1)
(0, 4)
(1, 0)
(1, 1)
(1, 2)
(1, 3)
(1, 4)
(3, 1)
(3, 2)
(3, 3)
Rectangle #2
(Start = Col_Offset + 4 = 0+4 = 4)
(End = Col_Offset + 0 = 4+0 = 4)
(Rectangle_Height
= 4)
(Yellow)
(Yellow)
(Brown)
(Brown)
(Pink)
(Brown)
(Pink)
(Pink)
(Pink)
(Pink)
Begin
d1 = Compressed_Image_Array[Iterator_Array_Index++];
If (d1 < 128)
Return d1;
Else
d2 = Compressed_Image_Array[Iterator_Array_Index++];
Return ((d1 - 128) * 256) + d2;
End If
End Function
Function GetAbsoluteColumnIndex - Get absolute value of Column Index
by adding with Column Offset
Begin
If (Compressed_Image_Array[Iterator_Array_Index] == 0xFF)
Itterator_Array_Index = Iterator_Array_Index + 1;
Return -1;
Else
Col_Offset = Col_Offset + GetNextValue();
Return Col_Offset;
End If
End Function
Function GetAbsoluteRowIndex - Get absolute value of Row Index by
adding with Row Offset
Begin
If (Compressed_Image_Array [Iterator_Array_Index] == 0xFF)
Itterator_Array_Index = Itterator_Array_Index + 1;
Return -1;
Else
Row_Offset = Row_Offset + GetNextValue();
Return Row_Offset;
End If
End Function
In
the
image
reconstruction
process,
the
Compressed_Image_Array has been read from start to end via
Iterator_Array_Index pointer without skipping or backwarding. So, the
algorithm complexity of reconstruction process is O(N) where N is
length of compressed image.
STEP 2 of the proposed method guarantees that the compressed
data size is not more than raw data size (N 3 x Image Size), so that
algorithm complexity is O(Image Size).
V. TEST RESULTS
The whole efficiency tests have been performed using STM32
CubeF7 Demonstration Application [13] on STM32F746G Discovery
Kit [14] with STM32F746NG microcontroller based on ARM CortexM7 core ( 1 Mbytes of flash memory and 340 Kbytes of RAM) with
4.3-inch 480x272 color LCD-TFT with capacitive touch screen (
Fig. 5). The main screen (Fig. 6 (a)) of demonstration application
consists with 8 image icons. Accessing each icon goes to various
screens (Fig. 6 (b-h)).
Each screen comprises of a number of drawing objects on a white
background. Those objects are divided into 3 types; polygon, text and
image (or icons). Polygons and texts could be drawn using sort of
various already built-in functions. However drawing images require all
image data to be stored in the internal memory as RGB format (3
bytes/pixels). All images in this test have been illustrated in red dashed
borders on each screen.
The testing started with measuring total pixel-count of all images
on each screen, later the number of clustering pixels on each screen
has been measured. Then, Grouping Ratio has be calculated. After
that, compassed data size and raw data size were measured. Finally
Compression Ratio has been evaluated. The testing results are as
listed in Table V.
VI. CONCLUSION
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
TABLE V.
[8]
Scre
en
All
Icons
#1
#2
#3
#4
#5
#6
#7
#8
Sum
29,335
19,650
24,310
44,400
47,880
51,402
62,400
19,600
298,977
Clusterin
g Pixel
20,924
4,404
21,711
40,601
46,527
47,991
58,789
13,767
254,714
a
Group
ing
Ratio a
(%)
71.33
22.41
89.30
91.44
97.14
93.36
94.21
70.24
85.20
Proposed
Method
32,178
4,770
11,904
17,064
6,127
13,660
13,959
22,683
122,345
Compr
ession
Ratio b
(%)
63.44
19.08
83.68
87.19
95.73
91.14
92.54
61.42
86.36
Compression ratio calculated by (Raw data size Proposed method data size) / Raw data size.
[9]
[10]
[11]
[12]
[13]
[14]
Suttichai Premrudeepreechacharn
I. INTRODUCTION
An Electricity Generation Authority of Thailand (EGAT)
is a state-owned enterprise which has responsible to generate
an electric power in Thailand since 1969. Now EGAT has
installed capacity more than 16,000 Mega Watt (MW) with
131 units of Generator Step-Up (GSU) transformers in power
plants. GSU transformer step-up the voltage and transmitted
electric power from generator to power transmission line
system as shown in Fig. 1. Lifetime of power plants usually
depends on the lifetime of main equipment such as boiler,
turbine, generator and GSU transformer. Approximately 80%
of GSU transformers in EGAT were aged over 15 years, more
than half life of power plants. Asset Life Cycle Management
of main equipment such as GSU transformer is a key success
of power plant reliability and availability. The remaining life
assessment is an important tool for lifetime management of the
power plants in EGAT.
[Source: http://www.bu.edu]
(2)
A t e
[Source: http://www.tbeaindia.com]
(3)
DPt
E
R (T 273)
1
DP0
k DP a
where A
R
T
E
t
(1)
where k = 0.0075
a = 1; for solvent polymer system
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
17.6
29.5
21.4
17.7
5.4
23.3
11.2
27.5
23.6
25.9
4.4
23.2
23.5
5.9
16.2
24.4
17.6
30.3
19.8
26.2
19.8
26.4
26.5
22.0
26.0
31.0
25.8
27.8
7.7
23.9
19.1
30.8
23.8
731
656
759
782
855
763
596
613
713
686
841
658
548
1028
766
629
939
679
676
567
775
454
596
654
734
720
717
785
959
898
729
792
807
GSU
Transformer
Lifetime
(years)
15.6
21.9
14.7
25.9
5.5
5.4
21.9
23.3
DP Value
767
744
781
671
866
749
872
699
35.26 264.131e
67592.46.DP-1.52
(4)
REFERENCES
[1]
ICESIT 2016
Entry
Faculty of Engineering