Full Chapter 3Rd Eai International Conference On Robotic Sensor Networks Rosenet 2019 Yujie Li PDF

3rd EAI International Conference on
Robotic Sensor Networks: ROSENET

2019 Yujie Li
Visit to download the full and correct content document:
https://textbookfull.com/product/3rd-eai-international-conference-on-robotic-sensor-ne
tworks-rosenet-2019-yujie-li/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
2nd EAI International Conference on Robotic Sensor

Networks: ROSENET 2018 Huimin Lu
https://textbookfull.com/product/2nd-eai-international-
conference-on-robotic-sensor-networks-rosenet-2018-huimin-lu/
Broadband Communications, Networks, and Systems: 10th

EAI International Conference, Broadnets 2019, Xi’an,
China, October 27-28, 2019, Proceedings Qingshan Li
https://textbookfull.com/product/broadband-communications-
networks-and-systems-10th-eai-international-conference-
broadnets-2019-xian-china-october-27-28-2019-proceedings-
qingshan-li/
Algorithms for Sensor Systems 15th International

Symposium on Algorithms and Experiments for Wireless
Sensor Networks ALGOSENSORS 2019 Munich Germany
September 12 13 2019 Revised Selected Papers Falko
Dressler
https://textbookfull.com/product/algorithms-for-sensor-
systems-15th-international-symposium-on-algorithms-and-
experiments-for-wireless-sensor-networks-algosensors-2019-munich-
germany-september-12-13-2019-revised-selected-papers-falko/
Interactivity Game Creation Design Learning and

Innovation 8th EAI International Conference ArtsIT 2019
and 4th EAI International Conference DLI 2019 Aalborg
Denmark November 6 8 2019 Proceedings Anthony Brooks
https://textbookfull.com/product/interactivity-game-creation-
design-learning-and-innovation-8th-eai-international-conference-
artsit-2019-and-4th-eai-international-conference-
Artificial Intelligence for Communications and
Networks: First EAI International Conference, AICON
2019, Harbin, China, May 25–26, 2019, Proceedings, Part
I Shuai Han
https://textbookfull.com/product/artificial-intelligence-for-
communications-and-networks-first-eai-international-conference-
aicon-2019-harbin-china-may-25-26-2019-proceedings-part-i-shuai-
han/
Artificial Intelligence for Communications and

Networks: First EAI International Conference, AICON
2019, Harbin, China, May 25–26, 2019, Proceedings, Part
II Shuai Han
https://textbookfull.com/product/artificial-intelligence-for-
communications-and-networks-first-eai-international-conference-
aicon-2019-harbin-china-may-25-26-2019-proceedings-part-ii-shuai-
han/
2nd EAI International Conference on Big Data Innovation

for Sustainable Cognitive Computing: BDCC 2019
Anandakumar Haldorai
https://textbookfull.com/product/2nd-eai-international-
conference-on-big-data-innovation-for-sustainable-cognitive-
computing-bdcc-2019-anandakumar-haldorai/
Security and Privacy in Communication Networks 15th EAI

International Conference SecureComm 2019 Orlando FL USA
October 23 25 2019 Proceedings Part II Songqing Chen
https://textbookfull.com/product/security-and-privacy-in-
communication-networks-15th-eai-international-conference-
securecomm-2019-orlando-fl-usa-october-23-25-2019-proceedings-
part-ii-songqing-chen/
Cloud Computing Smart Grid and Innovative Frontiers in

Telecommunications 9th EAI International Conference
CloudComp 2019 and 4th EAI International Conference
SmartGIFT 2019 Beijing China December 4 5 2019 and
December 21 22 2019 Xuyun Zhang
https://textbookfull.com/product/cloud-computing-smart-grid-and-
innovative-frontiers-in-telecommunications-9th-eai-international-
conference-cloudcomp-2019-and-4th-eai-international-conference-
EAI/Springer Innovations in Communication and Computing
Yujie Li
Huimin Lu Editors
3rd EAI
International
Conference
on Robotic
Sensor Networks
ROSENET 2019
EAI/Springer Innovations in Communication
and Computing
Series editor
Imrich Chlamtac, European Alliance for Innovation, Gent, Belgium
Editor’s Note
The impact of information technologies is creating a new world yet not fully
understood. The extent and speed of economic, life style and social changes
already perceived in everyday life is hard to estimate without understanding the
technological driving forces behind it. This series presents contributed volumes
featuring the latest research and development in the various information engineering
technologies that play a key role in this process.
The range of topics, focusing primarily on communications and computing engi-
neering include, but are not limited to, wireless networks; mobile communication;
design and learning; gaming; interaction; e-health and pervasive healthcare; energy
management; smart grids; internet of things; cognitive radio networks; computation;
cloud computing; ubiquitous connectivity, and in mode general smart living, smart
cities, Internet of Things and more. The series publishes a combination of expanded
papers selected from hosted and sponsored European Alliance for Innovation (EAI)
conferences that present cutting edge, global research as well as provide new
perspectives on traditional related engineering fields. This content, complemented
with open calls for contribution of book titles and individual chapters, together
maintain Springer’s and EAI’s high standards of academic excellence. The audi-
ence for the books consists of researchers, industry professionals, advanced level
students as well as practitioners in related fields of activity include information and
communication specialists, security experts, economists, urban planners, doctors,
and in general representatives in all those walks of life affected ad contributing to
the information revolution.
Indexing: This series is indexed in Scopus and zbMATH
About EAI
EAI is a grassroots member organization initiated through cooperation between
businesses, public, private and government organizations to address the global
challenges of Europe’s future competitiveness and link the European Research
community with its counterparts around the globe. EAI reaches out to hundreds of
thousands of individual subscribers on all continents and collaborates with an insti-
tutional member base including Fortune 500 companies, government organizations,
and educational institutions, provide a free research and innovation platform.
Through its open free membership model EAI promotes a new research and inno-
vation culture based on collaboration, connectivity and recognition of excellence by
community.
More information about this series at http://www.springer.com/series/15427

Yujie Li • Huimin Lu
Editors
3rd EAI International

Conference on Robotic
Sensor Networks
ROSENET 2019
Editors
Yujie Li Huimin Lu
Fukuoka University School of Engineering, Kyushu Institute
Fukuoka, Japan of Technology
Fukuoka, Japan
ISSN 2522-8595 ISSN 2522-8609 (electronic)

EAI/Springer Innovations in Communication and Computing
ISBN 978-3-030-46031-0 ISBN 978-3-030-46032-7 (eBook)
https://doi.org/10.1007/978-3-030-46032-7
© Springer Nature Switzerland AG 2021

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
We are delighted to introduce the proceedings of the 2019 European Alliance

for Innovation (EAI) International Conference on Robotic Sensor Networks
(ROSENET2019). The theme of ROSENET 2019 was “Cognitive Internet of Things
for Smart Society.” This proceedings highlights selected papers presented at the 3rd
EAI International Conference on Robotic Sensor Networks held in Kitakyushu,
Japan.
Today, the integration of artificial intelligence and Internet of Things has become
a topic of growing interest for both researchers and developers from academic
fields and industries worldwide. Artificial intelligence is poised to become the main
approach pursued in the next-generation IoTs research.
The rapidly growing number of artificial intelligence algorithms and big data
devices has significantly extended the number of potential applications for IoT
technologies. However, this also poses new challenges for the artificial intelligence
community. The aim of this conference is to provide a platform for young
researchers to share the latest scientific achievements in this field, which are
discussed in these proceedings.
The technical program of ROSENET 2019 consisted of 9 full papers from
24 submissions. Aside from the high-quality technical paper presentations, the
technical program also featured three keynote speeches. The three keynote speeches
were Prof. Hyoungseop Kim, Prof. Masuhiro Nitta, and Prof. Yuya Nishida from
Kyushu Institute of Technology, Japan.
Coordination with the steering chair, Imrich Chlamtac, was essential for the
success of the conference. We sincerely appreciate his constant support and
guidance. It was also a great pleasure to work with such an excellent organizing
committee team for their hard work in organizing and supporting the conference.
In particular, the Technical Program Committee, led by our Program Chair, Dr.
Shenglin Mu, Dr. Jože Guna, and Dr. Shota Nakashima, who have completed the
peer-review process of technical papers made a high-quality technical program. We
are also grateful to the Conference Manager, Lukas Skolek, for his support and
all the authors who submitted their papers to the ROSENET 2019 conference and
special sessions.
v
vi Preface
We strongly believe that ROSENET conferences provide a good forum for all
researchers, developers, and practitioners to discuss all science and technology
aspects that are relevant to Robotics and Cognitive Internet of Things. We also
expect that the future ROSENET conferences will be as successful and stimulating
as indicated by the contributions presented in this volume.
Fukuoka, Japan Yujie Li

Fukuoka, Japan Huimin Lu
Conference Organization
Steering Committee
Imrich Chlamtac University of Trento, Italy
Huimin Lu Kyushu Institute of Technology, Japan
Organizing Committee
General Chair
Huimin Lu Kyushu Institute of Technology, Japan
General Co-Chairs
Yujie Li Fukuoka University, Japan
TPC Chair and Co-Chair
Shenglin Mu Ehime University, Japan
Sponsorship and Exhibit Chair
Joze Guna University of Ljubljana, Slovenia
Local Chair
Tomoki Uemura Kyushu Institute of Technology, Japan
Workshops Chair
Guangxu Li Tiangong University
Publicity & Social Media Chair
Shota Nakashima Yamaguchi University, Japan
Publications Chair
Yin Zhang Zhongnan University of Economics and Law
Web Chair
Junwu Zhu Yangzhou University
Posters and PhD Track Chair
Quan Zhou Nanjing University of Posts and Telecommunications,
China
Preface vii
Panels Chair
Guangwei Gao Nanjing University of Posts and Telecommunications, China
Demos Chair
Hao Gao Nanjing University of Posts and Telecommunications, China
Tutorials Chairs
Csaba Beleznai Austrian Institute of Technology, Austria
Ting Wang Nanjing Tech University, China
Technical Program Committee

Dong Wang Dalian University of Technology, China
Weihua Ou Guizhou Normal University, China
Xing Xu University of Electronic Science and Technology of China,
China
Haiyong Zheng Ocean University of China, China
Rushi Lan Guilin University of Electronic Technology, China
Zhibin Yu Ocean University of China, China
Wenpeng Lu Qilu University of Technology, China
Yinqiang Zheng National Institute of Informatics, Japan
Jinjia Zhou Hosei University, Japan
Contents
Reinforcement Learning-Based Cell Intelligent Multimode

Frequency Reuse Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Zhiling Tang, Yuan Wang, Rushi Lan, Xiaonan Luo, and Simin Li
Image Registration Method for Temporal Subtraction Based
on Salient Region Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Suguru Sato, Huimin Lu, Hyoungseop Kim, Seiichi Murakami,
Midori Ueno, Takashi Terasawa, and Takatoshi Aoki
Extreme ROS Reality: A Representation Framework for Robots
Using Image Dehazing and VR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Akira Ueda, Huimin Lu, and Hyoungseop Kim
Double-Blinded Finder: A Two-Side Privacy-Preserving Approach
for Finding Missing Children . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Xin Jin, Shiming Ge, Chenggen Song, Xiaodong Li, Jicheng Lei,
Chuanqiang Wu, and Haoyang Yu
Complex Object Illumination Transfer Through Semantic and
Material Parsing and Composition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Xiaodong Li, Rui Han, Ning Ning, Xiaokun Zhang, and Xin Jin
Global-Best Leading Artificial Bee Colony Algorithms. . . . . . . . . . . . . . . . . . . . . . . 55
Di Zhang and Hao Gao
Intelligent Control of Ultrasonic Motor Using PID Control
Combined with Artificial Bee Colony Type Neural Networks . . . . . . . . . . . . . . . 71
Shenglin Mu, Satoru Shibata, Tomonori Yamamoto, Shota Nakashima,
and Kanya Tanaka
An Adaptable Feature Synthesis for Camouflage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Guangxu Li, Ziyang Yu, Jiying Chang, and Hyoungseop Kim
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
ix
Reinforcement Learning-Based Cell
Intelligent Multimode Frequency Reuse
Method
Zhiling Tang, Yuan Wang, Rushi Lan, Xiaonan Luo, and Simin Li
1 Introduction
Inter-cell interference exists in the upstream and downlink. Orthogonal Frequency

Division Multiple Access (OFDMA) is used in current and next-generation wireless
communication systems. Because signals in the cell are orthogonal to each other,
intra-cells interference is avoided [1]. However, users at the edge of a cell may use
the same time-frequency resource block to generate co-channel interference to the
neighboring cell, which is called inter-cell interference (ICI) [2]. As shown in Fig. 1,
user 1 in the cell and user 2 in the neighboring cell use the same frequency carrier
to transmit an uplink signal to the respective base station. The signal transmitted by
user 2 would interfere base station A, and the signal from user 1 to the base station A
would interfere base station B. These may seriously affect the signal-to-interference-
plus-noise ratio (SINR) which leads to a decline in communication quality. In Fig. 2,
there is the inter-cell downlink interference model. The base station A sends a signal
to user 1 and also receives the same frequency signal sent by the base station B. But
for user 1, it is also an ICI.
It can be seen from Figs. 1 and 2 that user 1 and user 2 are located at the edge
of the respective cells and vulnerable to be covered by the inter-cell interference
than users near the base station. This difference leads to the lower throughput of
the entire cell network. Moreover, there is inhomogeneity in the quality of service
(QoS) that the central users may be superior to the edge users. Therefore, improving
the QoS of the edge users by mitigating ICI is a future important task.
Frequency reuse (FR) is considered to be an effective method to solve ICI [3–5].
The frequency band is divided into several parts, which are respectively allocated
Z. Tang · Y. Wang · R. Lan () · X. Luo · S. Li

Guangxi Key Laboratory of Wireless Broadband Communication and Signal Processing, Guilin
University of Electronic Technology, Guilin, Guangxi, China
© Springer Nature Switzerland AG 2021 1

Y. Li, H. Lu (eds.), 3rd EAI International Conference on Robotic Sensor Networks,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-46032-7_1
2 Z. Tang et al.
Fig. 1 Uplink interference of

small interval interference
model
Fig. 2 Downlink
interference of small interval
interference model
Reinforcement Learning-Based Cell Intelligent Multimode Frequency Reuse Method 3
to the central users and edge users and ensure the edge users of the neighboring
cells using different frequency bands [6, 7]. According to the way of frequency
allocation, FR can be divided into static FR and dynamic FR. The static FR has
low complexity and low network signaling overhead and is easy to implement in
engineering. Dynamic frequency reuse method can dynamically modify frequency
reuse according to interference intensity, network load, network coverage, and other
conditions. But its disadvantage is that the network overhead is large and the
algorithm is complex.
In many cases, the reduction of interference is at the cost of the reduction of
spectral efficiency or energy efficiency. An advanced energy efficiency measure for
FR is proposed [8]. An adaptive partial frequency reuse scheme, which divides
a cell into two different virtual regions, is proposed [9]. With more deployed
femtocells to reuse available resources as much as possible, less resources can be
allocated to the portion of the macro-cell [10]. By classifying users by the principle
of avoiding major interference sources, different users use different reuse factors
to improve the throughput of the whole system [11]. To avoid interference and
crosstalk, an adaptive frequency reuse scheme for femtocell networks is used [12].
Exhaustive search and greedy descent methods are used to find the subcarrier and
power allocations in each cell and then are iteratively repeated among cells until a
predefined convergence criterion is satisfied [13]. The multilevel soft FR scheme
gives 2N power density limit levels, achieving a better interference pattern and
further improving the cell edge and the overall data rate [14]. In multiuser scenarios,
the service time of the network is minimized through the rational allocation of
resources [15].
In this chapter, an intelligent multimode FR scheme based on reinforcement
learning is proposed. Compared to traditional frequency reuse methods, this FR
method is more flexible for different wireless channel state. The reinforcement
learning-based scheme can enable optimal FR modes under different channel
conditions. Its goal is to improve the SINR of received signals and throughput.
2 Intelligent Multimode FR
FR is a concept proposed by Bell Labs in 1947, which is the cornerstone of cellular

mobile communication. FR factor represents the number of frequencies in a reuse
cluster. The larger the FR factor, the larger the reuse distance. Moreover, the analysis
shows that the larger the FR factor, the lower the spectrum efficiency and the smaller
the interference [16, 17]. Because the inter-cell co-channel interference decreases
with the increase of distance, the interference of the central users is smaller than that
of the edge users. The strategy of dividing central users and edge users is critical to
the performance of cellular networks.
In fact, the location of a user and the number of central users and edge users are
time-varying and random in the mobile cellular network. If the number of edge users
increases, and the FR mode with the FR factor of 3 is allocated, the requirements
4 Z. Tang et al.
of all users cannot be satisfied. Since a single frequency reuse cannot meet the
requirement, we propose an intelligent multimode frequency reuse method based
on reinforcement learning. Through reinforcement learning mechanism, one of the
four frequency reuse modes is selected to optimize network throughput, energy
efficiency, and spectrum utilization.
Firstly, the definitions of some main parameters that determine the mode of
frequency reuse are given. It is assumed that the state of the cellular network is
synchronized. The base station is in the transmitting state, and the user is in the
receiving state. ICI will be generated if users in the neighboring cell use the same
spectrum resource. The SINR is
P
SINR = (1)
(Iown + Iother ) + N
where P is the useful signal power received by the user, Iown is interference inside
the cell, Iother is interference between adjacent cells, and N is noise. P is defined as
follows:
P = Pbs + G − Ploss − Pshadow (2)
where Pbs is the transmit power, G is the gain of the antenna, Ploss is the path
loss, and Pshadow is the decline of shadow. Because OFDMA for downlink enables
orthogonality between symbols, the intra-cell interference is avoided. This leads to
the simplified SINR:
P
SINR = (3)
Iother + N
1. Throughput Kth
Throughput is related to the modulation method [18]. It is assumed that
Quadrature Phase Shift Keying (QPSK) modulation is adopted, and the coding
rate is 1/2. Then, throughput is defined as
n × Rb
Kth = (4)
Tf
where n is the number of subcarriers included in each resource block, Rb is

the bit rate of each subcarrier, and Tf is the frame length.
2. Coverage Kcover
If a SINR value is taken as a threshold γ , the percentage of the total users who
receive signals greater than or equal to γ is defined as coverage:
Number of users (γ ≥ 1)
Kcover = (5)
Total users
3. Spectrum efficiency Keff

The spectral efficiency is as follows:
Kth
Keff = (6)
B
where B is the channel bandwidth.
4. Spectrum utilization factor Uspe
In the FR scheme, the FR factor is defined as Kfr . The effective utilization of
the spectrum resources reflects the size of the reuse factor, which is defined as
1
n
Uspe = Pi (7)
S ∗ Pmax
i∈S
where S is the total number of sub-bands, Pmax is the maximum transmitted

power of all sub-bands, and Pi is the transmitting power that belongs to the
sub-band i. The soft FR factor is reciprocal to the spectrum utilization factor
as follows:
1
Kf r = (8)
Uspe
If the bandwidth of the edge region is 1/3 of the total bandwidth and the
transmitting power is Pmax , and the bandwidth of the central region is 2/3 of the
total bandwidth and the transmitting power is alpha αPmax (0 ≤ α ≤ 1), the following
equation can be obtained:

1 S 2S 1 + 2α
Uspe = Pmax + αPmax = (9)
S ∗ Pmax 3 3 3
It can be seen from Eqs. (8) and (9) that both the spectrum utilization rate and the
soft FR factor can be expressed by the transmission power ratio α. The larger the α,
the smaller the FR factor and the higher the spectrum utilization rate. According to
Shannon’s theorem, the capacity of edge users is Ce = (B/3)log2 (1 + SINRedge user ),
and that of center users is Cc = Blog2 (1 + SINRcenter user ). The FR factor of the
central user is 1, which can maximize the spectrum utilization efficiency, and Eq. (7)
shows that the throughput is maximized. Due to the serious interference to the edge
users, only by increasing the FR factor can SINR be enhanced, so the coverage area
of the cell has been greatly improved. However, according to Eq. (9), the spectrum
resources used are 1/3 of the total bandwidth of the system. When the throughput of
the central user is equal to that of the edge user, then
B
Blog2 1 + SI N R edge user = log2 (1 + SI N R center user ) (10)
3
6 Z. Tang et al.
Therefore,
SI N R edge user = (1 + SI N R center user )3 − 1 (11)
Equation (11) shows that the SINR of edge users must be larger than that of
central users under the same capacity. In the scheme of partial FR, the division of
center radius has a great influence on the system capacity of the cell.
According to the above analysis, four FR modes are proposed:
(a) The FR factor is equal to 1. The spectrum utilization rate is the highest, but the
edge users are seriously disturbed and the coverage rate is the lowest.
(b) Soft FR mode. The frequency band consists of high power band and low power
band. The high power band is twice the low power band.
(c) Soft FR mode. The frequency band consists of high power band and low power
band. The high power band is four times the low power band.
(d) The FR factor is equal to 1/3. Although the spectrum utilization rate is low
and the interference of the same frequency is small, the coverage rate becomes
larger.
Reinforcement learning neural network (RLNN) is used to select one of the four
frequency reuse modes which is most suitable for the current cellular network. For
the dynamic cellular environment, RLNN considers the SINR on its input, which is
called “Detection Neural Network” (DeNN), as shown in Fig. 3. DeNN is a com-
bination of multiple deep neural networks, and all deep neural networks are trained
in parallel by the same training data set, and finally, multi-objective performance
prediction values are generated. Each NN in DeNN has a feedforward structure
trained by Levenberg-Marquardt backpropagation algorithm [19]. It consists of
three fully connected unbiased layers: two hidden layers consist of 5 and 50 neurons
(weighted parameters of each neural network are 349), two layers use log-sigmoid
Detection Neural Network (DeNN)
NN0
Action ak NN1
,
, NNm 1
= ,
⋮ =1
SINR k
Fig. 3 Deep NNs used by reinforcement learning detection receive the same multidimensional
input
transfer function, and output layer uses standard linear transfer function in each
neuron. It uses the mean squared error (MSE) as the performance function under
two different early stop training conditions: the minimum error gradient is equal
to 10−12 , and the maximum validation checks is equal to 20. During the training
process, the training data sets are randomly divided into 70% for training, 15% for
testing, and 15% for validation, and all data are scaled to (0, 1). DeNN contains
20 NNs trained using the same training data set. The predictive output of DeNN is
used to classify input actions according to the performance threshold of FR mode
selector agent. The “action rejection probability” control randomly selects an action
from a good or bad set, which in turn is used to select FR mode.
3 Results
In system simulation, some of the parameters are adaptable. The range described
in Table 1 follows the sequence of operations described in FR control algorithm.
Multidimensional actions consisting of five parameters a = (MT;c; M; m; c) are
used in dynamic cell environment.
The simulation conditions are shown in Table 2.
We compared intelligent multimode FR with other FR methods and discussed
their advantages and disadvantages. In addition, the resource scheduling algorithm
Table 1 DeNN adaptable Parameter Variable Value range

parameters
Modulation type MT [QPSK,QAM]
Coding rate c [1/3,1/2,2/3,3/4,4/5]
Modulation order M [4,8,16,32]
Bandwidth B [2.5–10] MHz
FR factor FR [1–3]
Table 2 The simulation conditions

Parameter Value
Cell topology Macro cell
Number of cells 19
Station distance 1000 m
Number of cellular users 5000
Total number of resource blocks 50
BS transmit power 46 dBm
Fading type Large-scale fading, fast fading
Path model 128.1 + 37.6lg(d)dB d:km
Noise power spectral density −174.00 dBm/Hz
Number of slots 7
Data subcarrier of each resource block 12
8 Z. Tang et al.
CDF of the average received user SINR

1
mode 1
0.9
mode 2
0.8 mode 3
mode 4
0.7
0.6
CDF
0.5
0.4
0.3
0.2
0.1
0
-30 -20 -10 0 10 20 30 40 50 60
Average user SINR(dB)
Fig. 4 SINR distribution curve of four modes
plays an important role in obtaining higher throughput and fairness. It determines

the allocation method of wireless resources. Commonly used resource scheduling
algorithms are maximum carrier-to-interference ratio algorithm (MAX C/I), polling
algorithm (RR), and proportional fairness (PF) algorithm. Different resource alloca-
tion methods are validation in our simulations.
Firstly, four FR modes are simulated independently, and their SINR distribution
curves are shown in Fig. 4. As can be seen from Fig. 4, mode (a) has the highest
spectrum utilization, but the inter-cell interference is the most serious, and the
coverage is the worst. If the number of edge users is large, mode (b) should be
selected to achieve better performance. Mode (c) is appropriate if the user has a
lot of traffic and wants to get a lot of throughput. If the inter-cell interference is
minimum and the coverage area is large, mode (d) should be adopted.
The simulation results of FR based on RL are shown in Fig. 5. According to
the change of cellular environment, one of the four FR modes will be used in the
community. This proves that cellular can automatically select FR mode according
to the change of environment with reinforcement learning-based decision-making.
The effect of this method on network throughput is shown in Table 3. From Table 3,
we can see that the network throughput of intelligent FR method is significantly
higher than that of other FR methods, and it maintains a better coverage.
Finally, we simulate the signal interference characteristics based on the intelli-
gent multimode FR method, and the results are shown in Figs. 6 and 7. The statistical
level of each method is judged by different SINR. Cumulative distribution function
(CDF) is not the same under the same SINR condition. From Figs. 6 and 7, we can
3.5
3
mode
2.5
1.5
1
5 10 15 20 25 30
time
Fig. 5 The system switches between different modes
Table 3 Performance comparison of various FR methods

Mode name Throughput (Mbps) Coverage
Mode 1 12.8834 0.6751
Mode 2 14.3054 0.7456
Mode 3 16.9723 0.7986
Mode 4 9.8832 0.9562
Dynamic multimode frequency reuse 14.9742 0.8746
see that the SINR of our method for most edge users can be greater than 10 dB,
while other methods are mostly less than 10 dB. Therefore, the SINR of all users
and edge users has been significantly improved by using the FR method proposed
in this chapter.
4 Conclusion
The existing frequency reuse algorithms have some shortcomings; that is, they
cannot adjust the FR factor according to the change of the cell environment to
achieve the goal of network performance optimization. In this chapter, an intelligent
multimode FR method based on reinforcement learning is proposed to solve this
problem, which can automatically select the most appropriate frequency reuse mode
to improve the network throughput and coverage according to the real-time location
and traffic situation of the users in the cell. The simulation results show that the FR
method improves the throughput of the system and the coverage of the system.
10 Z. Tang et al.
1
Reuse factor 1
0.9
Intelligent multimode FR
SFR
0.8
0.7
0.6
CDF
0.5
0.4
0.3
0.2
0.1
0
-10 0 10 20 30 40 50
User average SINR value(db)
Fig. 6 Average SINR of all users
1
Reuse factor 1
0.9
Intelligent multimode FR
SFR
0.8
0.7
0.6
CDF
0.5
0.4
0.3
0.2
0.1
0
-10 -5 0 5 10 15 20
User average SINR value(db)
Fig. 7 Average SINR of edge users

References
1. L. Liu, T. Peng, P. Zhu, et al., Analytical Evaluation of Throughput and Coverage for FFR in
OFDMA Cellular Network, in IEEE Vehicular Technology Conference. IEEE, (2016)
2. D. Lee, G.Y. Li, S. Tang, Intercell interference coordination for LTE systems. IEEE Trans. Veh.
Technol. 62(9), 4408–4420 (2013)
3. G. Boudreau, J. Panicker, N. Guo, et al., Interference coordination and cancellation for 4G
networks. IEEE Commun. Mag. 47(4), 74–81 (2009)
4. A. Simonsson, Frequency reuse and intercell interference coordination in E-UTRA, in IEEE
Vehicular Technology Conference, Dublin, Ireland, (Sep. 2007), pp. 3091–3095
5. M. Rahman, H. Yanikomeroglu, W. Wong, Interference Avoidance with Dynamic Inter-
Cell Coordination for Downlink LTE System, in Wireless Communications & Networking
Conference, IEEE, (2009)
6. H. Mei, J. Bigham, P. Jiang, et al., Distributed dynamic frequency allocation in fractional
frequency reused relay based cellular networks. IEEE Trans. Commun. 61(4), 1327–1336
(2013)
7. A. Imran, M.A. Imran, R. Tafazolli, editors, A novel Self Organizing framework for adaptive
Frequency Reuse and Development in future cellular networks. Personal Indoor and Mobile
Radio Communications (PIMRC), in 2010 IEEE 21st International Symposium on; (2010) 26–
30 Sept. 2010
8. P. Zhu, T. Peng, Z. Liu, Energy Efficiency and Optimal Threshold for FFR chemes in
OFDMA Cellular Networks, in IEEE 10th International Conference on Signal Processing and
Communication Systems, (2016)
9. J.Y. Hwang, Y. Oh, Y. Han, Adaptive frequency reuse scheme for interference reduction in
two-hop relay networks, in Vehicular Technology Conference. IEEE, (2010)
10. S. Alotaibi, R. Akl. Dynamic Fractional Frequency Reuse (FFR) Scheme for Two-Tier
Network in LTE. in IEEE 8th Annual Ubiquitous Computing, Electronics and Mobile
Communication Conference, (2017), pp. 408–413
11. X. Fangmin, Fractional Frequency Reuse (FFR) and FFR-Based Scheduling in OFDMA
Systems, in International Conference on Multimedia Technology. IEEE, (2010)
12. G. Huang, J. Li, Interference mitigation for Femtocell networks via adaptive frequency
reuse[J]. IEEE Trans. Veh. Technol. 65(4), 2413–2423 (2016)
13. M. Qian, W. Hardjawana, Y. Li, Adaptive soft frequency reuse scheme for wireless cellular
networks[J]. IEEE Trans. Veh. Technol. 64(1), 118–131 (2015)
14. X. Yang, A multilevel soft frequency reuse technique for wireless communication systems.
IEEE Commun. Lett. 18(11), 1983–1986 (2014)
15. C. Yao, C. Yang, Z. Xiong, Energy-saving predictive resource planning and allocation. IEEE
Trans. Commun. 64(12), 5078–5095 (2016)
16. M. Belleschi, G. Fodor, D.D. Penda, Benchmarking practical RRM algorithms for D2D
communications in LTE advanced. Wirel. Pers. Commun. 82(2), 883–910 (2015)
17. M.N. Tehrani, M. Uysal, H. Yanikomeroglu, Device-to-device communication in 5G cellular
networks: Challenges, solutions, and future directions. IEEE Commun. Mag. 52(5), 86–92
(2014)
18. S. Deb, P. Monogioudis, J. Miernik, et al., Algorithms for enhanced inter-cell interference
coordination (eICIC) in LTE HetNets. IEEE/ACM Trans. Networking 22(1), 137–150 (2014)
19. L. Busoniu, R. Babuska, B.D. Schutter, D. Ernst, Reinforcement Learning and Dynamic
Programming Using Function Approximators (CRC Press, Boca Raton, FL, 2010)
Image Registration Method for Temporal
Subtraction Based on Salient Region
Features
Suguru Sato, Huimin Lu, Hyoungseop Kim, Seiichi Murakami, Midori Ueno,
Takashi Terasawa, and Takatoshi Aoki
1 Introduction
Cancer is the most common cause of death in Japan. It is reported that cancer causes
28.5% of death [1]. Bone is the most common site for metastasis in cancer and
is of particular clinical importance in breast and prostate cancers because of the
prevalence of these diseases. At postmortem examination, 70% of patients dying of
these cancers have evidence of metastatic bone disease. Carcinomas of the thyroid,
kidney, and bronchus also commonly give rise to bone metastases, with an incidence
at postmortem examination of 30–40% [2]. Bone metastasis causes pain, pathologic
fracture, hypercalcemia, myelosuppression, spinal cord compression, and nerve root
lesions Spinal cord compression is urgent for diagnosis and treatment because it
causes irreversible nerve palsy [3].
Computed tomography (CT) is frequently used as an image inspection method
in the current medical field. It contributes to early detection of bone metastasis and
evaluation of bone lesion size and bone stability. However, the number of images
has increased due to the improvement of the device performance, which is a burden
to the radiologist. In addition, inexperience radiologists have possibility of missing
small and haze lesions. Computer-aided diagnosis (CAD) system is expected to
contribute to solving these problems, and development has been advanced in recent
years. The CAD system is a diagnostic system in which a doctor considers computer
output such as detection and characterization of a lesion from a radiation image
based on computer vision as a second opinion. It is only used as a tool for providing
S. Sato · H. Lu · H. Kim ()

School of Engineering, Kyushu Institute of Technology, Fukuoka, Japan
e-mail: kim@cntl.kyutech.ac.jp
S. Murakami · M. Ueno · T. Terasawa · T. Aoki
University of Occupational and Environmental Health, Fukuoka, Japan

https://doi.org/10.1007/978-3-030-46032-7_2
14 S. Sato et al.
additional information to doctors and does not have to be equal to or greater than the
performance of a doctor. It needs to be complementary to the doctor’s ability [4].
Temporal subtraction technique is one of the fundamental technologies of CAD
system. It is a technique to generate images emphasizing temporal changes in lesion
region by performing a difference calculation between current image and previous
one of the same subject [5]. However, there is a problem that accurate temporal
subtraction images cannot be created due to deviation of the photographing position
by a difference in photographing time and body motion of the patient at the time
of photographing. Image registration is necessary to solve this problem. There
are many previous studies on image registration [6–8]. However, despite of the
urgent need to diagnose bone metastasis, there are few studies on image registration
focusing on vertebral region. Therefore, we propose an image registration system of
the vertebral region in this chapter.
The proposed method consists of three steps. First, we perform segmentation
of the region of interest (ROI) using position information of the spine based on
biology. Next, we select pairs of previous image and current image in which same
portion is depicted. Finally, we perform final image matching based on salient region
features (SRF) [9]. We apply the proposed method to 65 cases. We confirmed the
usefulness of temporal subtraction images for presenting the presence or absence
of bone metastasis. Section 2 shows details of our proposed method. Section 3
shows the performance evaluation of the proposed method with synthetic data and
an example of experimental results with real CT data. Finally, we conclude this
chapter in Sect. 4.
2 The Proposed Method
The procedure of the proposed method is shown in Fig. 1. In our method, we

performed noise removal by median filter and background removal using graph
cut which is one of region segmentation method. Next, we select pairs of images
by global image matching using center of gravity and normalized cross correlation
(NCC) and perform final image matching based on SRF. For more details, please
refer to the following.
2.1 Segmentation of Vertebral Region
We perform segmentation of the vertebral region as preprocessing for registration.

From the anatomical point of view (i.e., vertebral region is located at center of the
lower part on CT image), we set a rough region including the spinal region as a ROI.
We apply median filter to ROI to eliminate noise.
Image Registration Method for Temporal Subtraction Based on Salient Region Features 15
Fig. 1 The scheme of the

proposed method START
Segmentation
Global image matching
Final image matching
Create temporal subtraction image
END
2.2 Global Image Matching
In order to detect the difference of the CT slice in the body axis direction, we select
slices of previous image and current image of the same vertebral region is drown.
We calculate the position of the center of gravity of previous and current first slice
images. The center position on previous image Ip (i, j) is translated in a rectangular
area centered on the center of gravity position of current image Ic (i, j), and calculate
Ri that maximizes normalized cross correlation RNCC .

Ip (i, j ) Ic (i, j )
RNCC = (1)
Ip (i, j )2 × Ip (i, j )2
Rl = max RNCC (2)
We use the same operation for arbitrary number of previous images d while
updating slices of current image one by one and calculate maximum normalized
cross correlation Rl (l = 1, 2, · · · , d). The current image having the maximum value
among calculated Rl is selected as first pair. A slice on current image selected as
first pair is updated, and the maximum normalized cross correlation is calculated
with several slices before and after first slice of previous image. After that, slice
of current image is updated one by one, and all pairs are determined in the same
procedure. In global image matching, we need to pay attention to the intersection of
slice order when creating pairs.
2.3 Final Image Matching
We use conspicuity based on entropy and select SRF from vertebral region. Local
image matching based on SRF consists of three steps: Salient Region Feature
16 S. Sato et al.
Detection (SRFD), Region Component Matching (RCPM), and Region Configure

Matching (RCFM).
In SRFD, automatic detection of SRF necessary for later processing is performed.
We set the center coordinate X and the radius s. We calculate the entropy H(s, X)
based on the concentration probability function p(s, X) in the set circular area.
H (s, X) = − pi (s, X) log2 pi (s, X) di (3)

R
where i is a possible concentration value. After calculating the entropy at the center
coordinate X and the radius s, the radius is updated at the same center coordinate X,
and the entropy is recalculated. The radius is updated from 1 pixel to 10 pixels. The
radius value with maximum entropy value is defined as optimum scale Sx . Feature
value A(Sx , X) is expressed by optimum scales.
∂
A (Sx , X) = H · Sx · pi |Sx di (4)
R ∂s
The above process is the flow of SRFD at one center coordinate, and one feature
value is calculated for one center coordinate. Then, next center coordinate on the
image is set, and similar process is performed on the center coordinate. This is
performed on the entire image. In this chapter, the updated interval of the center
coordinates is set to 10 [pixel] in the X and Y directions. After calculating the
feature value at each center coordinate, these feature values are compared, and top
N number of circular regions is selected from the current image and the previous
image, where N = 20.
Next in RCPM, after automatically determining the pair P of SRF between
images based on similarity, we calculate the local likelihood Ll (P) of the pair. We
select one by one from the SRF of current image a and previous image b. We set
these as pair Pab and calculate the local likelihood Ll (Pab ) between selected SRF.
The entropy correlation coefficient (ECC) which is normalized mutual information
used for likelihood estimation is shown below:
Ll (Pab ) = maxECC (a, bθ ) (5)

θ
2H (a, bθ )
ECC (a, bθ ) = 2 − (6)
H (a) + H (bθ )
where bθ is a region b rotated at the rotation angle θ and H is binding entropy or

peripheral entropy based on the density values of the two regions. When likelihood
estimation is performed, region b is rotated every 3[deg] in the section of [−10, 10]
to maximize ECC and select top M pairs, where M = 30.
In RCFM, combination of pairs is performed incrementally using the top M pairs
selected by RCPM. We calculate global deformation amount Tg that transforms
reference image for registration with target image based on local deformation
amount Tl obtained from each pair. The movement amounts tx and ty are represented
by displacement of the center coordinate of SRF. Rotation amount is shown below:
θ̂ = arg maxECC (a, bθ ) (7)

θ
where θ̂ is the angle at which ECC obtains maximum value when SRF of the

previous image is rotated by angle θ . The likelihood Lg is used to evaluate the
similarity of the entire image. This evaluates using ECC between target image IF
and image IT obtained by transforming reference image IT and measures likelihood
of a pair combination composed by k pairs.

Lg Pa1 b1 ∩ · · · ∩ Pak bk = ECC (IF , IT ) (8)
When the pairs are incrementally combined, the pair showing maximum like-
lihood is selected as the pair to be selected first among M pairs. After selecting
the first pair, pairs are incrementally combined so that the global likelihood is
is an additional candidate pair. P = P
maximized. Pab j a1 b1 ∩ Pa2 b2 ∩ · · · ∩ Pak bk
is the latest pair combination. The global deformation amount Tg obtained from pair
combination Pj is shown below:
1
k
Tg = L li T li (9)
Lsum
i=1
where Lsum is the sum of each pair constituting a pair combination, Lli is likelihood
of the k-th pair of pair combination, and Tli is deformation amount of the k-th pair of
pair combination. After the first pair is determined, we select a maximum likelihood
pair from pairs not included in the latest pair combination as the additional candidate
. After a pair of additional candidate is selected, likelihoods are compared
pair Pab
in the
case ofcombining or not combining additional candidates. If the likelihood
Lg Pj ∩ Pab obtained by combining additional candidates is larger than the
likelihood Lg (Pj ) of current pair combination, additional candidate combinations
are performed and pair combination is updated to Pj = Pj ∩ Pab . If pairs of
additional candidates that are not included in the latest pair combination exist after
the pair combination has been updated, maximum likelihood pairs are selected from
them, and same processing is performed. If likelihood Lg Pj ∩ Pab obtained by
combining the additional candidates is lower than current likelihood Lg (Pj ) of pair
combination, or there are no pairs of additional candidates which are not included in
the latest pair combination, iterative process is terminated. The global deformation
amount Tg is calculated using the latest pair combination, and reference image is
deformed.
18 S. Sato et al.
3 Experimental Results
In this chapter, we evaluated registration accuracy in vertebral region using syn-

thesized data. An image of a certain vertebral region was defined as reference
image, and an image transformed by applying processing to the reference image
was defined as target image. The rate of overlapping region of the vertebral region
of target and reference image was defined as true positive (TP), and the rate of not
overlapping region (leakage region) was defined as false positive (FP). TP and FP
are shown below:
n (A ∩ B)
TP = × 100 (10)
n(A)
n(C)
FP = × 100 (11)
n(B)
where A is the spinal region of deformed target image, B is the spinal region of
reference image, C is the leakage region, and n is the number of pixels.
We used four synthetic data as target images: (i) the image obtained by rotating
reference image at rotation angle θ a = − 10, (ii) the image obtained by applying the
Gaussian filter to rotated image, (iii) the image obtained by adding artificial pseudo-
lesion region to rotated image, and (iv) the image obtained by adding random noise
of 5% to rotated image. Table 1 shows the performance evaluation, which is obtained
TP and FP of each target image. Furthermore, we compare the result of matching
by proposed method with the result of marking by radiologist.
Figure 2 shows three examples of matching result by proposed method and
marking result by radiologist. In Fig. 2, (a), (d), and (g) show temporal subtraction
images; (b), (e), and (h) show fusion images after matching; and (c), (f), and (i) show
marking images of lesion by radiologist. In fusion image, reference image is shown
in red, target image is shown in green, and overlapping region is shown in yellow
color. From the experimental results, abnormal area is enhanced on the temporal
subtraction images.
Table 1 Performance TP (%) FP (%)

evaluation
Rotation 100.0 12.16
Gaussian filter 70.40 0.00
Pseudo-lesion region 99.45 17.89
Random noise 83.05 16.95
Fig. 2 Example of experimental result. (a) Temporal subtraction. (b) Fusion. (c) Marking. (d)
Temporal subtraction. (e) Fusion. (f) Marking. (g) Temporal subtraction. (h) Fusion. (i) Marking
4 Conclusion
In this chapter, we proposed a method of image registration between current and

previous images in the vertebral region as a preliminary study to generate temporal
subtraction images from CT images and detected bone metastases on the vertebral
region. We evaluated performance using synthetic data. The image obtained by
rotating reference image at rotation angle θ_a = −10◦ gave TP 100.0% and FP
12.16%. The image obtained by applying the Gaussian filter to rotate image gave
20 S. Sato et al.
TP 70.40% and FP 0.00%. The image obtained by adding artificial pseudo-lesion

region to rotated image gave TP 99.45% and FP 17.89%. The image obtained by
adding random noise of 5% to rotated image gave TP 83.05% and FP 16.95%.
The proposed method succeeded in emphasizing the temporal changes and
contributed to improving the detection accuracy of the lesion part of doctor. This
deviation is due to the slice of same place that is not selected by global matching,
and only local deformation amount of the first pair is used as the global deformation
amount in RCFM step. In the future work, in order to improve the effectiveness of
the proposed method, we need to apply optimization using swarm intelligence [10,
11] and apply large deformation diffeomorphic metric mapping (LDDMM) [12] to
the data we used for comparative experiments.
Acknowledgement This work was supported by Leading Initiative for Excellent Young
Researchers of the Ministry of Education, Culture, Sports, Science and Technology-Japan
(16809746) and JSPS KAKENHI Grant Number 16K14279, 17K10420.
References
1. Ministry of Health, Labor and Welfare, Vital statistics, http://www.mhlw.go.jp/toukei/saikin/

hw/jinkou/suikei16/index.html
2. R.E. Coleman, Clinical features of metastatic bone disease and risk of skeletal morbidity. Clin.
Cancer Res. 12(20), 6243–6249 (2006)
3. G.J. Cook et al., Detection of bone metastases in breast cancer by FDG PET: Differing
metabolic activity in osteoblastic and osteolytic lesions. J. Clin. Oncol. 16(10), 3375–3379
(1998)
4. K. Doi, Computer-aided diagnosis in medical imaging: Historical review, current status and
future potential. Comput. Med. Imaging Graph. 31(4–5), 198–211 (2007)
5. K. Doi et al., Computer-aided diagnosis in radiology: potential and pitfalls. Eur. J. Radiol.
31(2), 97–109 (1999)
6. Z. Zhong et al., 3D-2D deformable image registration using feature-based nonuniform meshes.
BioMed Res. Int. 2016 (2016)
7. A. Sotiras et al., Deformable medical image registration: A survey. IEEE Trans. Med. Imaging
32(7), 1153–1190 (2013)
8. B. Rodriguez-Vila et al., Methodology for registration of distended rectums in pelvic CT
studies. Med. Phys. 39(10), 6351–6359 (2012)
9. X. Huang et al., Hybrid image registration based on configural matching of scale invariant
salient region features. CCVPRW 11, 167–179 (2004)
10. D. Karaboga, An Idea Based on Honey Bee Swarm for Numerical Optimization, Technical
Report-TR06, Erciyes University, Engineering Faculty, Computer Engineering Department,
(2005)
11. X.S. Yang, Cuckoo Search via Levy Flights, in World Congress on Nature & Biologically
Inspired Computing, IEEE Publications, (2009), pp. 210–214, arXiv:1003.1594v1
12. R. Sakamoto, Temporal subtraction of serial CT images with large deformation diffeomorphic
metric mapping in the identification of bone metastases. Radiology 285(2), –629, 639 (2017)
Extreme ROS Reality: A Representation
Framework for Robots Using Image
Dehazing and VR
Akira Ueda, Huimin Lu, and Hyoungseop Kim
1 Introduction
Japan is known as a country that is likely to suffer natural disasters such as

typhoons, heavy rains, heavy snow, floods, landslides, earthquakes, tsunamis, and
volcanic eruptions worldwide. Although the area of Japan’s land is only 0.28%,
20.5% of earthquakes of magnitude 6 or more that happened in the whole world
occurred in Japan, and 7.0% of active volcanoes in the whole world exist in
Japan. In addition, 0.3% of disaster deaths worldwide is recorded in Japan, and
11.9% of the damage amount received in the disasters of the whole world is
the damage amount in Japan [1]. Japan is more likely to suffer from natural
disasters than other countries. In addition, artificial disasters such as accidents at
construction sites (e.g., gas explosion accidents in tunnels) are also increasing. The
deterioration of infrastructure such as tunnels, expressways, and dams is a problem,
and complex accidents caused by aging infrastructure such as corrosion are on the
rise. Responding to such equipment accidents is also a subject [2].
As the threat of natural disasters and artificial disasters increases, the importance
of preparing for them increases. In terms of disaster prevention, it is important
not only to take measures after the disaster such as emergency response, recovery,
reconstruction, measures to prevent disasters, and inspection and maintenance of
infrastructure and equipment [2].
It is always dangerous for a person to work in disaster response and recovery
where it is difficult to do work. Disaster prevention measures and inspection and
maintenance of social infrastructure and equipment are similar. In order to solve
these problems, introduction of robots under dangerous work environments is
A. Ueda · H. Lu () · H. Kim

School of Engineering, Kyushu Institute of Technology, Fukuoka, Japan
e-mail: dr.huimin.lu@ieee.org

https://doi.org/10.1007/978-3-030-46032-7_3
Another random document with
no related content on Scribd:
rocks, mates,” says I, trying to speak stern, though, Lord love yer, I
felt for all the world as if I was agoin’ to choke.
We left two men to do sentry-go, and set to work to overhaul them
rocks and all the ground and underwood which was near ’em; but not
a blessed sign of anything could we find except the marks of feet,
and in places these were plentiful enough, but ’tis a most tremenjous
difficult thing to track any one in a forest where there’s always a lot of
dead leaves and such like muck about.
Well, I’d be afraid, and that’s the truth, to say how many hours we
spent over that there hopeless job of searching for our lost
shipmates. We scoured the forest in all directions, made our way into
the valley to look for the enemy’s camp, but found nothing but a
burned-out fire; and overhauled the surrounding cliffs—and there
was a pretty few of ’em, mind you—in hopes of lighting upon some of
the blooming caves we’d heard so much talk about. It just was talk,
and nothing else—that I’ll take my affidavy to. Where were those
swabs o’ mutineers and the confounded Creole niggers that we’d
been sent after? and where was their hidden cargo? Shiver my
timbers, if I wouldn’t have slugged the lot of ’em if I could have got
the chance. I don’t like fallin’ out with people as a rule, specially
strangers; but when it comes to their kidnappin’ your own shipmates,
and p’raps cutting their throats, or giving ’em foul play of some sort
or another, why then my monkey gits up, I can tell you, and I’d think
it child’s play to corpse the lot of ’em if they came within reach of my
cutlass.
When the sun began to go down, I knew well enough that we must
knock off and make tracks for the camping-ground that Mr.
Thompson had spoken of. The alarm must be given as quickly as
possible, so that the whole force might turn-to and join in the search;
but then a fresh difficulty came to the fore, for we’d lost our guide,
and hadn’t the remotest idea of the bearings of this here bivouac.
However, I don’t want to make a long yarn out of this part of the
business, so I’ll only tell you that we made a sternboard, so to speak,
and found our way back with uncommon difficulty to the rough cart-
track we had quitted early in the afternoon. Here, of course, we got
on the spoor of the brigade, and had no difficulty in tracking ’em to
their camping-ground. ’Twas a weary march, and we were footsore
and a bit done up when we got there, which was after dark; and we
ran a narrow shave of being shot down by our own sentries.
Well, naterally there was a tremenjous sensation amongst all
hands when they heard our terrible story. The gunnery lootenant was
almost beside himself, and vowed vengeance on them dastardly
rebels when he caught them. He wanted to set off at once, but on
coming to think it over calmly, he saw that ’twas wellnigh impossible
till the morning. In the first place, every mother’s son was tired out by
the long stretch they’d had in full marching order; and, secondly, it
was that dark—there being no moon—that it would have been
wellnigh impossible to find our way through the forest, to say nothing
of the chances it would give them thundering swabs of mutineers
and niggers to go for us under cover of the darkness. They couldn’t
have drubbed us, of course, but through their sarpentine cunning
they might have harassed us like old winky.
The orders was passed round therefore to start in the morning, at
the first break of dawn.
Through the adventures what follered it seemed to me as I ought
to make some alterations in this here gimcrack of a log; but when I
consults the gunnery orficer—“You just leave it as it is, Jim
Beddoes,” says he; “’tis more interestin’,” he says. Well, ’tis a new
thing to me that anything I should have written, being no scholard,
should be interestin’; but, there, I’ve had my say. The yarn is spun,
such as it is, and I can only say in conclusion, and with many
apologies for takin’ of the liberty, “My dooty to you, and wishing you
all good-luck.”
CHAPTER XII.
TAKEN PRISONER.
W HEN I found myself blindfolded, gagged, and securely bound

by the ruffians who had sprung upon me, my heart sank within
me, for I fully thought that my fate was sealed. And what an awful
fate it might be I hardly dared to contemplate. I had heard, of course
—and the remembrance recurred to me now with additional force—
that the mutineers of the Flying-fish, had cruelly and in cold blood
murdered their captain and chief mate, and thrown their bodies to
the sharks. Men who had committed such a deed of infamy were
capable of anything—even of torturing us.
Till I was completely exhausted, I fought desperately with these
fiends, and so, as I mentioned before, did the gunner and Ned
Burton. I must say this much for our captors, that they carried out the
affair with amazing cleverness, coolness, and audacity; for our
seizure was accomplished in such a manner as to ensure secrecy
and obviate any necessity for using weapons.
I really felt more dead than alive when I was ruthlessly dragged
away from the spot where the struggle had taken place. I can assure
the reader that it is no joke to be knocked about and roughly handled
by four active, muscular rascals who have received orders to secure
you at any cost.
Feeling bruised, giddy, and faint therefore, I had to submit to the
indignity of being dragged away like a dead pig, convinced that an
awful fate most surely awaited me and my unfortunate companions.
What would I not have given to see the brave Jim Beddoes and
his trusty men rushing down like a whirlwind upon these audacious
scoundrels, and making them fly in all directions. It was terribly
tantalizing to think of the well-armed little detachment of stanch
friends actually within hail, if we could only have managed to shout
to them for assistance. I felt extremely sorry for Mr. Triggs, for I knew
what his feelings must have been, and how he must have blamed
himself for having led us into danger.
In spite of my weak and terribly anxious state, I could not help
wondering how the foe had got wind of our approach, and been able
to make such elaborate preparations for our capture.
For about ten minutes these ruthless fellows dragged us at their
heels, and then came to a sudden halt, and I heard them exchanging
remarks in a language which sounded to me like bad Spanish. I
concluded, therefore, that none of the mutineers of the Flying-fish
were amongst them, which struck me as a peculiar circumstance.
Whilst revolving this in my mind, I felt that some one was carefully
examining the bandage over my eyes, as if to ascertain whether I
could see or not. The next moment I was hurried forward by the
united exertions of several men, and felt myself being borne through
what felt like a plantation of young trees or shrubs; and shortly
afterwards realized the very disagreeable fact that I was being
impelled down a flight of rude, slippery steps, on which I had
considerable difficulty in keeping my footing. If it had not been for the
iron grip of my jailers, I should have been precipitated from the top to
the bottom. To my bewildered mind this rude staircase seemed
interminable, and at every step my spirits sank lower and lower, and
at length, when we appeared to reach the bottom, they were pretty
well at zero.
A burst of coarse laughter here fell upon my ears, and some
snatches of a Spanish song, mingled with which, like a humming
accompaniment, was the sound of falling water.
I made a great effort to pull myself together and manfully face my
fate whatever it might be; but my nerves, I must confess, had been
terribly shaken, and my success was only very partial.
I was determined, however, that I would not flinch from death if our
murder had been determined upon, as I felt that it would be casting a
dishonour upon our nation not to meet even that terrible fate with
calm courage—hard, very hard as it seemed, to be cut off at such an
early age, when life was at its brightest epoch.
Having now arrived at the bottom of the steps, I was forced along
a smooth sandy floor; my conductors laughing and joking among
themselves in an uproarious manner, evidently in the highest spirits,
and not afraid of being overheard.
The truth flashed across me in a moment. I was in a cave, deep
down under the surface of the ground!
The bandage over my eyes was now removed and the gag taken
from my mouth; but the lashings which secured my arms were, I
need hardly say, not interfered with. With very unnecessary force,
and before I had had time to look around me, I was pushed down
into a corner of the cave, and two swarthy, cut-throat-looking rascals
with loaded pistols in their hands were told off to guard me securely.
The first thing these fellows did was to rob me of my silver watch and
chain, and the small amount of money I had about me.
It was an impressive and weird tableau which met my gaze as
soon as my eyes had got accustomed to the semi-gloom of the
dimly-lighted cavern, the feeble illumination of which evidently came
from natural fissures in the rocky roof, many of them being half-
choked with vegetation.
This cave, in which I found myself a most unwilling prisoner, was
vast in size and unequal in shape; but I could not see its whole
extent from the spot where I lay. The sound of falling water was loud
and continuous, and I fancied that there must be a cascade—
invisible to me—falling through an orifice in the roof into a pool
beneath. Notwithstanding this fact, the cavern appeared to me to be
very dry.
My first impulse, of course, was to look for Mr. Triggs and Ned
Burton. The former I could not see, and concluded that he had been
placed in one of the invisible portions of the cave; but my coxswain
was sitting leaning against the rough wall on the opposite side of our
subterranean prison, a very angry and perturbed expression upon
his bronzed face. Three armed men closely watched his every
movement. Our eyes met, but we did not dare even to make a sign
to each other for fear of the consequences.
As near as I could tell, there were about a score of Creole
desperadoes grouped about in the cave. They were all well-armed,
truculent-looking ruffians, and were, to my intense astonishment,
arrayed in a sort of semi-nautical costume, consisting of a striped
jersey, white-duck trousers, and sombrero hats, whilst some of them
sported gay-coloured kerchiefs knotted around the throat. They were
all young men, with the exception of one individual, who at this
moment was seated upon a barrel smoking a cigar. He was a man of
about fifty, with a fine physique and clearly-cut features; but the
expression of his face was cruel, stern, and unbending. He wore a
moustache and beard, the latter streaked with grey; and between his
eyes some perpendicular lines had been deeply impressed. The
dress of this chief—for so I deemed him to be—differed somewhat
from that of his followers, as he wore a silk shirt fastened with
elaborate studs and with a turn-down collar affixed to it, and a thin
blue flannel pea-jacket with plain gilt buttons. He wore no hat, and
his hair was curly, dark, and thick, except at the temples, where it
was a little worn and tinged with grey. His weapons consisted of a
handsome sword and a pair of beautifully-chased silver-mounted
pistols; whilst a short cavalry carbine—evidently of the very best
make—was reclining against the barrel on which he was seated. At
his feet lay a large, fierce-looking Cuban bloodhound, possessed of
muscular limbs and cruel-looking fangs. I could not resist shuddering
when I looked at this animal, for I felt sure that at a sign from its
master it would make short work of any one who had displeased
him.
When my eyes first lighted upon this chieftain, of anything but
prepossessing looks, I found that he was examining with great
curiosity the weapons which had been taken from us by his
followers, and his mind seemed specially exercised by the
mechanism of the service-revolvers. I remember distinctly that
although I was at the time fully expecting to be led away somewhere
to be summarily executed, yet I felt nervous at the careless way the
chief was handling weapons which were loaded in all their cylinders,
and with the working of which he could have but a very slight
acquaintance, if any.
There was a very large assortment of arms stored in every nook
and cranny, and hung around the cavern’s rude walls, consisting
chiefly of various kinds of swords, scimitars, and daggers. There was
also a trophy of pistols and tomahawks, and another of boarding-
pikes; whilst a goodly number of rifles and muskets were ranged in
rough stands occupying two sides of the cave. The area of the latter
appeared to be, as nearly as I could judge, fifty feet in length by thirty
in width; but these measurements did not include, of course, those
portions which were invisible to me. I could detect no exit from the
spot where I lay except the rude steps by which we had descended,
but I felt convinced that there was an outlet elsewhere.
Two or three of the chief’s followers were engaged in handing our
appropriated weapons to him, and giving him—at least so I
presumed—an account of our capture. The remainder of the gang
were grouped about in various attitudes in different parts of the cave:
some sprawling on mats or blankets spread carelessly on the sandy
floor, and engaged in smoking and playing cards; others eating and
drinking with ferocious appetites; whilst a few more industrious ones
were polishing their weapons, or mending tattered clothes and boots,
singing or whistling to themselves the while. I could see no signs of a
fire or of cooking utensils, or, indeed, of any provisions. The men
who were eating appeared to be devouring sausages and coarse
bread, which they drew from their capacious pockets, and a strong
and very disagreeable odour of garlic pervaded the close stagnant
atmosphere of the cave. I was terribly thirsty, and the rushing sound
of the waters of the invisible cascade very much aggravated the
feeling. I longed to bury my burning lips in the sparkling clear pool
that lay, I felt sure, beneath it.
Then the remembrance of the story of the valuable cargo of the
Flying-fish flashed across my memory. According to the reports of
the spies and guides, all this valuable booty which the mutineers had
seized upon was stowed away in caves well known to their insurgent
allies. Well, here was a cave big enough in all conscience to hold the
cargo of an ordinary merchantman; but where in the name of all that
was wonderful was this much-talked-of loot? I could see no traces of
it, and I felt sure that my eye embraced the main portion of the giant
cave. Stay! there might be other and smaller caverns opening from
this one, in which the spoils might have been stowed for greater
safety. I felt sure that this was the solution of the mystery, and longed
to go on a voyage of exploration.
Oh that distressing feeling of burning thirst! I at length summoned
up enough courage to say in an imploring tone to one of my
guardians, “Agua, agua!”
To my surprise the fellow fetched a pannikin of water and held it to
my parched lips.
Oh how delicious, how cooling, how like nectar was that heavenly
draught!
I was obliged to stop at length with a sigh of gratitude, and murmur
“Gracias.”
A slight feeling of hope animated me after this incident, for it struck
me that if our instant execution had been decided upon, the very
scantiest consideration would have been shown for our wants—
indeed that they would not have been considered in any one
particular.
That we had fallen into the hands of consummate scoundrels went
without saying, of course. A glance at the cruel, repellent visage of
the chief, or at those of his swarthy, stick-at-nothing myrmidons, was
sufficient to settle that question.
Who were these mysterious individuals?
The more I looked at them, the more I felt convinced that they
were not Creole insurgents. At any rate, they were not the same men
whose fort we had stormed and whose forces we had dispersed only
a day or two previously. Neither did they resemble—except in their
olive-coloured complexions—the dashing irregular cavalry which had
met with the same disastrous fate. The men into whose clutches we
had fallen were much more of the sailor type, and this impression
was not given by the cut of their clothes alone. It was also betrayed
by the cut of their jibs, as sailors express it—that is, by their
physiognomies having an unmistakable nautical stamp upon them.
Had the Cuban insurgents any vessels under their orders—
privateers or cruisers of any kind? I could not answer the question,
though I was aware of the fact that the Spanish navy in the West
Indies was deplorably weak.
Ned Burton was emboldened by my successful application for
water to ask for some too. It seemed to me that his guards went
rather grudgingly to fulfil his request. However, I don’t suppose Ned
much minded that, so long as he got the water. The enormous
quantity he drank without so much as winking actually provoked
laughter among the swarthy gang who were watching him.
The rope lashings which bound my arms together were
exceedingly painful, and seemed like the cruel fangs of animals
gnawing into the flesh. I only knew about half a dozen Spanish
words, but I utilized these as well as I could in an endeavour to
persuade my guards to relax the tightness of my bonds. Having been
successful in one instance, I hoped to be in another. Vain hope! The
only result of my appeal was a violent kick from one of the sullen
sentinels, which made me heartily wish that I had held my peace.
A burst of ribald laughter rang through the gloomy cavern, echoing
on the rocky roof, and drowning for the moment the roar of the
cascade.
I saw an angry frown gather upon Ned Burton’s brow, and a flush
mount into his cheeks, and felt sure that he was itching to feel his
fingers around the throat of my assailant. It would have gone hard
with the latter if my brawny coxswain had been free from his
lashings. As it was, he had to chafe under the heavy yoke of
enforced inactivity, and I had to do the same.
It must have been shortly after this that, in spite of my strange
surroundings, I fell fast asleep, and was lost in the world of happy
dreams.
It was the sleep of exhaustion, of tired-out nerves, of a bewildered
brain. Nature, kindly dame, comes to our rescue on these occasions;
and deeply grateful ought we to be to her for her loving care of us.
I do not know at all how long I had been slumbering in the arms of
Morpheus, when I was awakened by a hideous uproar—by a perfect
babel of angry, excited shouts and vindictive yells.
With my head in a whirl, and scarcely realizing where I was, I
strove to rise to my feet. At that instant, however, I felt the cold
muzzle of a pistol pressed against my temples, and heard a torrent
of angry invective from one of my sentinels.
CHAPTER XIII.
NED IS FLOGGED.
T HE cold barrel of the pistol quickly brought me to myself; but I

felt strangely bewildered. The glare of many torches was in my
eyes, and was reflected on the irregular roof and walls of the giant
cave with a strange and weird effect. These flickering rays of light
also glinted with an almost dazzling brightness on the muskets and
other weapons that lay about, and fell in a more subdued fashion on
the fierce-looking visages of about a score of the inmates of the
cavern, who had unconsciously assumed highly-dramatic attitudes.
All had sprung to their feet, and some had drawn pistols from their
belts, whilst others waved the bright blades of swords over their
heads.
The chief did not appear to be present, as far as I could tell in the
confusion, nor did I see his savage-looking dog; but one individual,
whose face seemed strangely familiar to me, was standing
somewhat apart from the others. He was talking loudly in a high-
pitched voice, and seemed to be in a very excited state of mind,
using a great deal of gesticulation, and pouring forth what sounded
uncommonly like the language of menace.
Where had I previously seen that swarthy, unprepossessing face,
those cruel eyes, that tall ungainly form?
As I racked my brains in an effort to solve the mystery, the angry
and energetic orator suddenly paused in his copious flow of
invective, and in a dramatic manner pointed straight at Ned Burton.
At the same moment every eye and every weapon was turned
upon my coxswain in a terribly ominous manner.
The truth flashed across me in an instant, and made me feel sick
at heart.
This angry orator, this passionate denouncer, was none other than
the cruel mule-driver who, on the way up from the sea-coast, had
been flogged by Ned, in accordance with the instructions of the
gunnery lieutenant.
Could any meeting have been more unfortunate?
My coxswain’s conduct during this trying ordeal was beyond all
praise. That he recognized the man I had not the smallest doubt. He
could not have failed to do so.
By the flickering light of the torches, which were jammed into rude
sockets in the wall, I could see that Ned was sitting on the floor of
the cave in the exact spot where I had seen him before, and watched
by the lynx eyes of three armed desperadoes. He looked quite
unconcerned at the angry tirade of his accuser; but I noticed that he
kept his eyes steadily fixed upon those of the mule-driver, and that
the latter seemed rather uneasy under this steadfast gaze. I could
not help feeling alarmed as to the outcome of this episode and its
accompanying disorder, for I felt convinced that Ned’s enemy was
exaggerating the details of his story in the hopes of taking an ample
revenge for his fancied wrongs.
I looked about for Mr. Triggs, but could not see him anywhere. It
was evident that my former supposition was correct, and that he was
confined in another portion of the cavern. The worthy gunner could
have been of no possible use to us in this emergency, but I felt that it
would have been a comfort to see him, and to feel assured of his
presence and sympathy.
As I was gazing in a sort of fascinated way at the strange scene, a
hush fell as if by magic on the excited assembly, a loud resonant
voice rang throughout the cavern, and I saw the chief’s form
appearing in the distance. He had evidently been absent for some
little time, and was unaware of the advent of the mule-driver.
His followers fell back abashed as he strode forward imperiously
into the centre of the cavern, closely followed by the big bloodhound.
I fancied that the autocrat’s eyes flashed angrily, and that there was
an unusual flush mantling on his olive-tinted cheek.
There was now a dead silence, and the only sound that broke the
stillness was the rush and plash of the cascade. Even this appeared
to annoy the chieftain, for he called one of his followers to his side
and gave him some directions in sharp, curt tones. The man saluted,
and quickly disappeared around a corner of the cavern. After a
minute or two had elapsed the sound of the falling water abruptly
ceased. The stream had evidently been turned into another channel.
The chief now called up and interrogated the mule-driver, who had
meanwhile slunk aside as if afraid of the reception he was likely to
meet with from the perturbed leader of the gang.
How earnestly I wished at that moment that I had been acquainted
with even a smattering of the Spanish language! Even Ned was a
better scholar in this respect than I was, and I noticed that he bent
forward eagerly to listen.
The mule-driver was evidently only too ready to be questioned, for
he immediately stepped forward and launched out once more into
the tale of his wrongs. The chief moodily listened, with bent brows
and fingers toying idly with the butts of his silver-mounted pistols.
Once or twice he glanced vindictively at my coxswain, but during the
greater part of the rehearsal kept his sombre eyes fixed stonily on
the cavern floor.
As for Ned, not a muscle of his face moved whilst the acting of this
little drama was being carried out. His clear brown eyes were fixed
intently on the speaker, though from time to time they rested for a
moment on the strongly-marked features of the listening chieftain. I
do not think I have mentioned before that my coxswain was a good-
looking fellow of the best sailor type. His age at this time must have
been about thirty-four or thirty-five, and though he was only of
medium height, his build was that of a very active and athletic man.
Beards were not in vogue in the navy at the time of my story, and as
Ned had no personal fancy for whiskers, his face remained clean-
shaven.
The mule-driver gradually worked himself up into a passion as he
progressed in his second recital, and at the end thereof had recourse
to a very dramatic proceeding; for as he wound up his long-winded
peroration with many emphatic words, he tore off a light cotton jacket
he was wearing, and exposed to view his bare shoulders, on which
several weals were distinctly visible.
This action had the desired effect, as no doubt the wily mule-driver
had foreseen. The inmates of the cave rose en masse to their feet in
an excited manner, and shouted and yelled like a lot of maniacs. But,
as had happened before, the chief did not at all approve of this
hubbub, and promptly and indignantly quelled it. Doubtless he was
afraid of his followers’ cries being heard outside the cave by
unfriendly ears.
A sickly dread took possession of me. What was to be the next act
in the drama? As to that, I was not long left in suspense.
The chief, as soon as quiet reigned again, summoned to his side
the mule-driver and four of his own satellites. The result was a
conference which did not last very long. At its conclusion the mule-
driver looked perfectly radiant, not with genuine pleasure, but with
gratified malice. He cut various capers on the sandy floor of the
cave, and snapped his bony, yellow fingers at Ned Burton. Indeed, I
do not know to what extremities he might not have gone if the chief
had not sternly called him to order.
At a signal from that autocrat, six of his followers advanced quickly
and in a determined manner towards Ned Burton, exchanging jocose
remarks as they did so.
The seaman looked perfectly impassive. He might have been a
statue sculptured out of marble for all the notice he took of what was
going on around him.
In a moment these six ruffians had seized Ned and forcibly raised
him to his feet, their victim offering no resistance, though I fancied I
saw the corners of his firmly-cut lips quiver for a moment.
My heart began to beat tempestuously. What were these
malevolent-looking rascals about to do to my coxswain? Their looks
boded him no good; that much was certain.
Meanwhile the mule-driver had disappeared from the scene. He
now, however, came skipping back into view in a jaunty manner,
carrying in his hands a long, cruel-looking cane.
Then the awful truth flashed across me—Ned was to be flogged!
I think I lost my head for a few moments, and it is difficult for me to
record my actions with any exactitude. I only know that in spite of the
lashings which confined my arms, I bounded to my feet and
attempted to rush madly forward to my coxswain’s rescue. I believe I
raved, and stormed, and entreated mercy, all in the same breath.
The only thing I do most distinctly remember was seeing Ned’s
handsome pale face turned in my direction with an imploring
expression upon it. It was this which brought me to myself, and
showed me the utter futility of my Quixotic attempt at interference.
It was the feeling of devotion which I had for my coxswain, of
course, that prompted the action; but fortunately I realized in time
that any rash effort on my part might do him more harm than good.
I was about to stagger back to my corner, sick at heart, when one
of the sentinels who had been detailed to watch me, and was
enraged at my attempt to elude him, rushed up and knocked me
down with a heavy blow from the butt-end of a pistol.
Thank God! I became unconscious, and therefore did not witness
the cruel flogging to which Ned was subjected.
CHAPTER XIV.
IN THE PIRATES’ CAVERN.
S OME considerable time must have passed before my senses

returned to me. When I opened my eyes again in a scared way,
wondering whether or not I was in the midst of an extremely
disagreeable dream, I found myself lying on my back on the floor of
the cavern; which latter seemed strangely dim and dark, rendering it
very difficult to distinguish anything. My head ached terribly, and I
think I groaned aloud as the remembrance of late events came
crowding into my mind.
“Hush!” was suddenly uttered in low friendly tones coming out of
the semi-darkness around me; “be as quiet as possible, and I’ll try to
give you some water.”
With astonishment and relief I recognized the voice of Mr. Triggs.
What was he doing here, and where were all our captors gone?
The cave seemed buried in silence as well as in gloom.
I endeavoured to raise myself, and quickly realized that my arms
were still confined by the cruel rope lashings. Like arrows through
my temples shot agonizing neuralgic pains. Was my brain on fire? It
felt uncommonly like it. My throat and lips, too, were dry and parched
with feverish thirst, and my tongue seemed to cleave to the roof of
my mouth.
And now the gunner’s friendly face came peering through the
darkness as he bent over me.
“Keep up your pecker, Mr. Darcy,” he said soothingly; “’tain’t no
use giving way. There is a pannikin of water here close to your head,
and I want you to screw yourself round a bit and try to get a drink
from it. I can’t assist you, worse luck, for my arms are bound too.”
Water! The very notion brought back with a rush the vitality which
was at such a low ebb in my system. I was dying for a drink of water.
It seemed to cost me little or no effort now to raise myself, and I
turned over upon my side and recognized the pannikin of which the
gunner had spoken. Crawling over the short space which separated
me from it, I plunged my face into the cool refreshing liquid and
lapped greedily like a dog.
It was fresh life to me as far as actual strength of body was
concerned, but it did not abate the pains in my head, which were still
intense.
The gunner was kneeling near me. I could now distinguish objects
more distinctly, but the only source of illumination in the cave, as far
as I could see, was one flickering sputtering torch, and this was
dimly flaring away at the farthest extremity of the great rock-
chamber. The ghostly light fell weirdly on Mr. Triggs’s strongly-
marked features and crouching figure. I noticed that he was looking
pale, worn, and anxious. When he saw me glance at him, however,
he tried to smile encouragingly.
My eyes roamed around the great deserted cavern with a puzzled
expression. Surely it was all a dream! Where was the gang of
shouting and yelling desperadoes, the chief, the mule-driver? Ha!
where was my brave and faithful coxswain, Ned Burton? When I had
last seen him—
The gunner divined my thoughts.
“I must explain everything hurriedly,” he said in a low rapid voice,
“for at any moment that swab of a chief and his men may return.
First with regard to Ned Burton. I’m thankful to say the brave fellow
lies sleeping over yonder in the dark shadow of that projecting
corner. Don’t you see him?”
I glanced hurriedly at the spot indicated by Mr. Triggs, and now
that my eyes had got more accustomed to the peculiar light,
managed to discern a dark, huddled-up, human form.
“But were not those cruel brutes going to flog him?” I falteringly
asked; “or was it only the beginning of a dreadful dream?”
“They did flog him, poor chap,” answered the gunner, in tones
which shook with emotion; “but Ned bore it like a man. Not a sound
escaped his lips during the punishment, and he did not faint. I’ve
been to him and given him some water, and he is now happily
asleep. Of course his back is in a terrible state, but he declares it will
be all right in a few days. I feel sure, however, that he is very much
exhausted.”
“What a mercy that he can sleep!” I said. “But tell me, Mr. Triggs,
what has become of our captors?”
“They were hastily summoned away by a messenger,” answered
the gunner, “a man who seemed breathless and nearly exhausted,
as if he had run for a very long distance without stopping to rest. The
whole gang then hurriedly seized their weapons, extinguished all the
torches except the one you see there, and rushed tumultuously out
of the cave by an entrance which was visible to me where I lay. Half
a dozen men, however, remain on guard just outside, and may enter
the cave at any moment. They have not done so yet, however, and
are, I think, in too excited a state to pay much attention to us,
besides—”
“Then can’t we escape?” I asked impetuously. “Is there no other
exit—the way we came in, for instance?”
The gunner shook his head dolefully.
“Whilst our arms are bound we’re helpless,” he answered; “if they
were only free, we’d make a dash for liberty. I’ve hurriedly examined
the flight of steps by which we entered, and I’ve found that the
opening to the outer air is securely closed by an iron door. You may
depend upon it that every possible precaution against our escaping
has been taken by these villains, for they’ve got their wits about ’em.”
“I expect that Beddoes and the rest of the detachment are on their
trail,” I said, “and that is what caused them such alarm. Perhaps we
shall be rescued sooner than we think.”
“I hardly think so,” said the gunner dubiously, “for I’m certain that
the messenger had travelled a long distance.”
“Then perhaps the main body have got an inkling that something
has gone wrong,” I suggested; “Beddoes may have sent a message
to Mr. Thompson asking him to push up his forces.”
“There hasn’t been time for all that, Mr. Darcy, when you come to
think of it. I shouldn’t say it was more than nine o’clock now, if so
much. No, ’tis a mysterious business altogether, and I can’t make
head or tail of it, and that’s the truth.”
“You don’t think, Mr. Triggs, that they mean to murder us?”
“Make your mind easy about that, lad. The ghastly deed would
have been done before this, I reckon, if they had resolved upon foul
play. What their little game is I can’t for the life of me say. ’Tis a plot
I’d like to unravel, I can tell you, but I’m all in the dark—all in the
dark.”
“I don’t believe that they’re insurgents at all,” I said; “they look for
all the world like a gang of—”
“Mum’s the word!” interrupted Mr. Triggs, scrambling nimbly to his
feet; “I hear the swabs coming back.” And so saying he took to his
heels, and in a moment had disappeared around a projecting part of
the cavern wall.
I heard the subdued buzz of many voices speaking in the distance,
and felt a cold shiver go down my back. Then came the sound of
many shuffling footfalls, the clang of weapons, and the louder tones
of an excited troop of men. The guard that had been stationed
outside rushed in to see that their prisoners were safe, and finding
that we were in our respective corners, ran out again to meet their
returning comrades. A few seconds later the whole gang came
straggling into the cavern, threw themselves on the ground, and
began talking rapidly in what I believed was a patois peculiar to
themselves. The chief did not appear to be with them.
A few minutes later I was thankful to see some preparations made
for supper, for I was ravenously hungry, and should have eaten
heartily of a monkey pie or devilled crocodile at that moment. No fire
was lit, but from some recess in the cave various comestibles were
brought forth and placed on upturned cases and barrels. A few more
torches had been lit, and though they were not enough to illuminate
the cave properly, I could see distinctly what was going forward.
At this moment the chief, carrying his carbine over his shoulder,
stalked in, followed closely by his immense dog, the latter—for some
reason best known to himself—snarling and showing his long white
fangs, whilst saliva dropped continuously from his heavy jowls. I had
a mortal dread of this savage-looking animal, for I felt instinctively
that if ever by good-fortune we should be enabled to make our
escape from our present captors, the bloodhound would be put upon
our trail to hunt us down. The reflection was a very disturbing one,
and I strove to banish it, but in vain. Certainly no plan of escape
seemed feasible, but a British midshipman is sanguine even when
things look irretrievably dark, and, like the immortal Mr. Micawber,
trusts that eventually something favourable will turn up.
The chief strode into the centre of the cavern and flung himself
down on a pile of goatskins, his dog taking up a position at his feet,
and keeping its eyes fixed upon those of its master. Its snarling now
ceased, much to the relief of my overwrought nerves.
A man hurried forward to bring the leader some refreshment, and I
was astonished to recognize in this servitor the mule-driver who had
been the cause of Ned’s flogging. The reason I was astonished was
that this fellow had completely changed his costume, and now wore
the same nautical dress as the chief’s immediate followers.
I divined the truth at once. The cunning rascal had been told off to
spy upon the movements of the naval brigade, and had enlisted as a
camp-follower for that very purpose. On being relieved of his duties
he had doubtless found some means of following us—himself
unseen—and had been the actual moving spirit in the underhand
arrangements for our capture; for he evidently owed allegiance to the
chief, and was perfectly familiar with the life of the cavern.
Fresh sentries had been told off to guard us, and one or two of
these fellows went off to procure us some food. By this time I was
nearly fainting for want of nourishment. I glanced across to Ned’s
corner, and fancied that he had been awoke by the noise in the cave,
but I could not tell for certain.

Full Chapter 3Rd Eai International Conference On Robotic Sensor Networks Rosenet 2019 Yujie Li PDF

Uploaded by

Copyright:

Available Formats

You might also like

Full Chapter 3Rd Eai International Conference On Robotic Sensor Networks Rosenet 2019 Yujie Li PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Full Chapter 3Rd Eai International Conference On Robotic Sensor Networks Rosenet 2019 Yujie Li PDF

Uploaded by

Copyright:

Available Formats

3rd EAI International Conference on

Robotic Sensor Networks: ROSENET

2nd EAI International Conference on Robotic Sensor

Broadband Communications, Networks, and Systems: 10th

Algorithms for Sensor Systems 15th International

Interactivity Game Creation Design Learning and

Artificial Intelligence for Communications and

2nd EAI International Conference on Big Data Innovation

Security and Privacy in Communication Networks 15th EAI

Cloud Computing Smart Grid and Innovative Frontiers in

More information about this series at http://www.springer.com/series/15427

3rd EAI International

ISSN 2522-8595 ISSN 2522-8609 (electronic)

© Springer Nature Switzerland AG 2021

We are delighted to introduce the proceedings of the 2019 European Alliance

Fukuoka, Japan Yujie Li

Technical Program Committee

Reinforcement Learning-Based Cell Intelligent Multimode

Inter-cell interference exists in the upstream and downlink. Orthogonal Frequency

Z. Tang · Y. Wang · R. Lan () · X. Luo · S. Li

© Springer Nature Switzerland AG 2021 1

Fig. 1 Uplink interference of

FR is a concept proposed by Bell Labs in 1947, which is the cornerstone of cellular

P = Pbs + G − Ploss − Pshadow (2)

where n is the number of subcarriers included in each resource block, Rb is

3. Spectrum efficiency Keff

where S is the total number of sub-bands, Pmax is the maximum transmitted

SI N R edge user = (1 + SI N R center user )3 − 1 (11)

Detection Neural Network (DeNN)

Table 1 DeNN adaptable Parameter Variable Value range

Table 2 The simulation conditions

CDF of the average received user SINR

Fig. 4 SINR distribution curve of four modes

plays an important role in obtaining higher throughput and fairness. It determines

Fig. 5 The system switches between different modes

Table 3 Performance comparison of various FR methods

Fig. 6 Average SINR of all users

Fig. 7 Average SINR of edge users

S. Sato · H. Lu · H. Kim ()

© Springer Nature Switzerland AG 2021 13

2 The Proposed Method

The procedure of the proposed method is shown in Fig. 1. In our method, we

2.1 Segmentation of Vertebral Region

We perform segmentation of the vertebral region as preprocessing for registration.

Fig. 1 The scheme of the

Global image matching

Final image matching

Create temporal subtraction image

2.2 Global Image Matching

2.3 Final Image Matching

Detection (SRFD), Region Component Matching (RCPM), and Region Configure

H (s, X) = − pi (s, X) log2 pi (s, X) di (3)

Ll (Pab ) = maxECC (a, bθ ) (5)

where bθ is a region b rotated at the rotation angle θ and H is binding entropy or

θ̂ = arg maxECC (a, bθ  ) (7)

In this chapter, we evaluated registration accuracy in vertebral region using syn-

Table 1 Performance TP (%) FP (%)

In this chapter, we proposed a method of image registration between current and

θ̂ = arg maxECC (a, bθ ) (7)