Motion History Images for Action Recognition and Understanding

132 Pages·2013·2.793 MB·English

by  Md. Atiqur Rahman Ahad (auth.)

SpringerBriefs in Computer Science
Series Editors
Stan Zdonik
Peng Ning
Shashi Shekhar
Jonathan Katz
Xindong Wu
Lakhmi C. Jain
David Padua
Xuemin Shen
Borko Furht
V. S. Subrahmanian
Martial Hebert
Katsushi Ikeuchi
Bruno Siciliano
For furthervolumes:
http://www.springer.com/series/10028
Md. Atiqur Rahman Ahad
Motion History Images
for Action Recognition
and Understanding
123
Md.AtiqurRahman Ahad
Department of AppliedPhysics, Electronicsand
Communication Engineering
Universityof Dhaka
Dhaka
Bangladesh
ISSN 2191-5768 ISSN 2191-5776 (electronic)
ISBN 978-1-4471-4729-9 ISBN 978-1-4471-4730-5 (eBook)
DOI 10.1007/978-1-4471-4730-5
SpringerLondonHeidelbergNewYorkDordrecht
LibraryofCongressControlNumber:2012950858
(cid:2)Md.AtiqurRahmanAhad2013
Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation,broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionor
informationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purposeofbeingenteredandexecutedonacomputersystem,forexclusiveusebythepurchaserofthe
work. Duplication of this publication or parts thereof is permitted only under the provisions of
theCopyrightLawofthePublisher’slocation,initscurrentversion,andpermissionforusemustalways
beobtainedfromSpringer.PermissionsforusemaybeobtainedthroughRightsLinkattheCopyright
ClearanceCenter.ViolationsareliabletoprosecutionundertherespectiveCopyrightLaw.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexempt
fromtherelevantprotectivelawsandregulationsandthereforefreeforgeneraluse.
While the advice and information in this book are believed to be true and accurate at the date of
publication,neithertheauthorsnortheeditorsnorthepublishercanacceptanylegalresponsibilityfor
anyerrorsoromissionsthatmaybemade.Thepublishermakesnowarranty,expressorimplied,with
respecttothematerialcontainedherein.
Printedonacid-freepaper
SpringerispartofSpringerScience+BusinessMedia(www.springer.com)
This book is dedicated to—
my wife Shahera Hossain (Ruba) &
my lovely daughter Rumaisa Fatima!
Foreword
It is commonly observed that what used to be science fiction eventually becomes
engineering fact. This is certainly the case for human–computer interaction
technologies, where speech recognition and synthesis, gesture recognition, face
recognition, facial expression analysis, and other technologies that were intro-
duced to many people in sci-fi settings have become, or are in the process of
becoming, a part of our daily lives. These transitions are not the result of magic
potion or of mad scientists alone in their labs, but of many researchers proposing
andtestingmethodsthat,bitbybit,bringthetechnologiesclosertoreality.These
communitiesofresearchersarelargelydrivenbywhatispossible,whatisneeded,
and what is interesting.
The topic of human behavior analysis has received agreatamountofattention
in recent years, due to the increasing computational power available to analyze
video streams as well as the interest in applications of action and gesture recog-
nition in security and surveillance, human–computer interaction, gaming, and
otherareas. MotionHistoryImages(MHIs),firstintroducedbyDavisandBobick
in the 1990s, provide certain advantages in representational power and efficiency
thathaveprovedtobequiteusefulintherecognitionofhumanmovement,action,
and activity. In addition to its effective practical use, there is an elegance in the
MHI representation that makes the method a valuable tool for learning and
thinking about human movement recognition.
Though still a young researcher with a long, promising career of contributions
ahead of him, Md. Atiqur Rahman Ahad has a substantial background of experi-
ence in optics, image processing, and computer vision with a focus on human
activity analysis, developing and assessing methods for representing, and recog-
nizing a range of spatio-temporal events. With a solid track record of relevant
conference and journal publications, Ahad is well positioned to reintroduce the
conceptofMotionHistoryImagestoperhapsanewgenerationofresearchersand
system builders, who—armed with considerably better tools than in the 1990s—
mayusethisasastartingpointtopushtheboundariesofwhatispossibleinhuman
activity analysis and perhaps in other areas of motion-based video interpretation.
Ahad provides an overview of the MHI method and discusses variations of the
vii
viii Foreword
method,technicalchallengesthatstillneedtobeaddressed,arangeofapplications
for which the technique is well suited, and datasets available to researchers
working in the area. The tutorial level approach makes this book accessible to
students new to the area, which also provides a useful summary for experienced
researchers in the field.
I encourage readers to approach the book with the question in mind, how can
webuildonthis?Thatis,ifweunderstandthestrengthsandlimitationsoftheMHI
approach to representing and recognizing activity, we are better prepared to
explore the follow-up methods that may make robust, real-time human activity
analysiscommonplacetomorrow.Itispossible;itisneeded;and,asthisbookmay
convince you, it is certainly interesting.
Santa Barbara, CA, August 2012 Matthew Turk
Preface
This book is about the Motion History Image (MHI) and its developments.
Broadly, it falls in the arena of computer vision and image processing. On a
smaller scale, the book is on action/activity/gesture/gait recognition and human
behavior analysis. It is known that human action analyses and recognition are
challenging problems due to large variations in human motion and appearance,
camera viewpoint, and various other issues related to environment settings.
Thoughthefieldofactionandactivityrepresentationandrecognitionisrelatively
old, yet it is not well understood by the student and research communities. Some
important but common motion recognition problems are even now unsolved
properly by the computer vision community. However, in the past decade, a
number of good approaches have been proposed and evaluated subsequently by
many researchers. Among those methods, MHI method and its variants get sig-
nificant attention from many researchers in the computer vision field due to their
better robustness and performance.
This book attempts to reduce the gap of information and materials on com-
prehensive outlook—through various strategies from scratch to mainly the MHI
method and its variants, challenges, and applications. This book targets the stu-
dents and researchers who have knowledge on image processing at a basic level
and would like to explore more on this area and do research. The step-by-step
methodologies will encourage one to move forward for a comprehensive knowl-
edgeoncomputervisionforrecognizingvarioushumanactions,withtheaidofthe
concept of the motion history images and its variants.
Some good features of this book are:
• No book is available as per my knowledge that covers action recognition con-
sidering the MHI in a comprehensive manner in the arena of computer vision.
• Bothstudents(whohavebasicknowledgeonimageprocessing)andresearchers
can use this book for their academic learning and research purposes.
• The book covers the MHI—a state-of-the-art method on recognition approach
and it can guide a fellow on how to move forward.
ix
x Preface
• This book can be used as a textbook (full or partial) for students of level-2 or
above,whohavestudiedbasicimageprocessingalreadyinanearliersession—
both for undergraduate and graduate students.
The book comprises four chapters. Chapter 1 has background on action recogni-
tion. Chapter 2 presents some aspects of action recognition. As the book is tar-
getedtobeashortenedone,notmuchdetailshavebeencoveredinthesechapters.
In Chap. 3, the core chapter of on the Motion History Image method and its
variants, detailed analysis, and variants are covered. This chapter illustrates on
how to advance from one method to another, how to improve an existing method
and mitigate its shortcoming, how to employ an approach in various other appli-
cations, etc. Therefore, this chapter is highly beneficial for an enthusiastic
researcherorstudenttomoveforward.Finally,Chap. 4coverssomedatabasesthat
are employed for the MHI method. Due to page limits, detailed issues are not
coveredbutreferredforfurtherreading.Ihopethatthisbookwillhaveanimpact,
serve the community in a positive direction, and help people to learn.
Foranyqueryonthisbookorcorrections,pleasewritetome. Iamthankfulto
my teachers, co-researchers, and students. All images are produced, except some
of these which are taken with the kind permission of Springer Science+Business
Media B.V. A few images are created by Upal Mahbub. I am grateful to Ahmed
Boudissa for his kind time to check the book. I specially thank Simon Rees of
Springer for his constant support to finish the book.
Thank you all!
Fukuoka, Japan, August 2012 Md. Atiqur Rahman Ahad
Email: [email protected]
Preferred websites:
http://IJEI.org
http://ICIEV.org
http://IJCVSP.com
http://benjapan.org/IJE
http://benjapan.org/ICEAB
Acknowledgments
Iammuchgratefultomysensei—SeijiIshikawa,KyushuInstituteofTechnology,
Japan.IamthankfultoJ.K.Tan(KyushuInstituteofTechnology,Japan),H.Kim
(Kyushu Institute of Technology, Japan), Z. H. Mahmood (University of Dhaka,
Bangladesh),andotherteacherswhohelpedmethroughoutmytenureasastudent.
A great number of researchers, academicians, and students were behind the
scenetomotivatemetowritethisbookandIamthankfultothemfromthecoreof
my heart. I thank Upal Mahbub, of Bangladesh University of Engineering and
Technology, Bangladesh who produced a few images for the book. I am thankful
tomyparents—nonecanreplacethem,theirefforts.Parentsaretremendous!Ican
never forget the efforts of my beloved mother who single-handedly managed and
supported after my father died at an early age. I am ever grateful to her.
Finally, Shahera Hossain—to whom I am extremely thankful for her support
and encouragement. She has always supported me even when she was in hospital
and passing through bad patches.
xi