Woontack Woo, Ph.D.
KAIST UVR Lab.
Daejeon, Korea
ISO SC24 WG9 Meeting @ Seoul, Korea, 01/19/2017]
Based on Two Korean Standards
TTAK.KO-10.0434 (Dec. 2010): Representation for AR: POI, Content, Link
TTAK.KO-10.0851 (Dec. 2015): Metadata for AR content visualization & sharing
Proposal on “Metadata: data about data”
Metadata representation for context-aware MAR services
what::Status, (PoI ID, physical/virtual object ID)
who::Role, (author profile)
when::Time, (created, modified)
where::Location, (geographical/relative position)
how::Action, (link/interaction/response method)
why::Goal, reason
Meta-tagging: the process of adding metadata with 5W1H to various
physical/virtual assets/entities
Metaverse for MAR Services
Modeling
Interaction
Networking
Sensing
Metaverse
Measured
Space
Physical
Space
Virtual
Space
SensingModeling
Anchoring
(Augmentation)
UVR, Context-aware MAR for DigiLog World (Woo)
3D Link with IoE (inc. IoT) between dual (real & virtual) spaces with
additional information
Augmentation of Context-of-Interest, not only sight but also
Everything (sound, haptics, smell, taste, etc.)!
Bidirectional Interaction/collaboration in dual spaces
SPACES
Private (3rd skin)
Social
General/Public
Real space
IoT/IoE, SNS
Seamless
Augmentation btw
dual spaces
How
to
Link
Seam
lessly? CoI
Ancho
r
MAR-
Content
Virtual space
Indoor MAR Applications [UVR Lab 2008-10]
Outdoor MAR Applications
Holistic Layers : In addition to 3D CG
.
.
.
.
.
.
IoT/IoE
Real/Physical
Virtual/Cyber
Information/
Knowledge
SPACE
TIME
Layers Social wisdom
Increased Interest and Demands on MAR
Popularized since the advent of “smartphone”
Increasing # of MAR Apps
However, Incompatible MAR Contents
Most AR contents are not re-usable in other applications
Due to different “processes, data structure/format, etc.”
In addition, differences in describing PoI & linking approach
So, The key of Context-aware MAR services is
Metadata representation for “Virtual, Real & Anchoring”
However, no standard “metadata” Format for MAR content
Now, What do you think …
What should we do for useful MAR services?
MAR
Services
…
DB B
DB CDB D
DB E
Metadata: Structured Data Fields with 5W1H
(Who, When, Where, What, How) + Why
Representation of Metadata Should be
Interoperable: platform independent
Ca: Semantic searching, filtering, and sharing of content
ASAP: be concise, compact, yet flexible, extensible to use
Scope: Metadata for refering
MAR content (image, text, video, audio, 3D model)
PoI: real entity (person/object/space/location) or content
MAR Anchor: Link between RC & VC:– contains the
instructions for MAR visualization
ISO/IEC CD 18039
External 
Services
User profile 
(Experience, 
Knowledge, 
Behavior)
Simulator 
(Event, Space)
Renderer
MAR Execution Engine
Contextualized PoIs
Personal Device Display
User 
Interface
MAR Scene
(Events, Targets, …)
Environment (or 
wearable) Sensors 
for Real World 
Capturer
Internet of Thing 
(IoT)
Sensors
Sensors
GPS
Compass
Gyro sensor
Personal Device 
Embedded
Camera
Media 
Assets
MAR Content
Virtual
Object
Physical 
World
Open API
Physical World
User
User
Input
User
Command
Enhanced
Experience
MAR 
Visualization
Recognition Data
Context Analyzer 
PoI Analyzer
User Localization
View Point Setting
PoI Analyzing
Context Processor
Context Integration
Context Inference
Context Management
Recognized
PoIs
User Context
Sensor Data 
PoI Recognition
Data
MAR 
Anchor
Prope
rties
MAR Ontology
MAR
Entity
MAR
Entity Info.
Metadata Type Definition
Who.Creator String Person or group that have rights or custody on the object
Who.Owner String Person or group that create the object
When.Created Time DateTime The time when the object was created
When.Modified Time DateTime The time when the object was modified
Where.PoI Location GPS GPS coordinate of the PoI
Where.PoI Direction Float
Direction of front of the PoI measured by Compass. If the PoI i
s view-point invariant(i.e. symmetric) then the value is NULL
Where.PoI 3D
Coordinate
Float (x, y
(or, z))
Position of MAR Contents in MAR scene (Can be 2D or 3D)
Where.GPS Location
of Recognition Data
GPS
GPS coordinate of the Recognition Data when the data are
acquired
Where.Contents 3D
Coordinate
Float (x, y
(or, z))
Position of MAR Contents in MAR scene
Where.Contents 3D
Direction
Float (x, y, z)
3D normal vector of the Contents in MAR scene. Absolute
value of normal vector is 1.
Where.User Location GPS GPS coordinate of the user
Where.User Direction Float (x, y, z) 3D value of device pose measured by gyroscope of user device
Metadata Type Definition
What.PoI ID String Identification of the PoI
What.Recognition Data ID String Identification of the Recognition Data
What.Recognition Data FileFile Local data file path or URL for Recognition Data
What.Contents ID String Identification of the Contents
What.Contents type String Type of the Contents
What.Contents Reference String References of the Contents
What.Contents Data File File Local data file path or URL for Contents Data
How.Lens Intrinsic
Parameter of Acquisition
Device
Float (fx,fy,
cx,cy,k1,k2,k3,r)
Intrinsic parameter of the embedded lens of capturing devic
e. fx, fy are focal length of the lens. cx, cy are principal point
of the lens. (c.f. Skewness of the cell array is 0.) k1,k2,k3,r ar
e distortion parameters of lens acquired by camera calibratio
n. Those parameters can be converted with respect to camer
a model. (In this case, Pin-hole model)
How.Lens Intrinsic
Parameter of User Device
Float (fx,fy,
cx,cy,k1,k2,k3,r)
Intrinsic parameter of the embedded lens of user device
How.Contents Scale
Float (x, y
(or , z))
Size of the MAR contents in MAR scene. (dimension of the
scale depends on the content)
MAR Ontology
Devised to cover the
Relation-based data
model
Consists of correlated
MAR entities and
properties to describe
and Literals Value for
the primitive value
All object in MAR
Ontology has the URI
(Unified Resource ID)
instead of ID
Through the PoI, user
can start the exploring
the MAR ontology
MAR
Entity
Literals
Value
Literals
Value
MAR
Entity
MAR
Entity
Property
Property
MAR Ontology
Property
PoI
MAR
Contents
MAR
Entity Info.
Property
MAR
Contents
MAR
Entity Info.
MAR
Entity Info.
MAR
Anchor
MAR
Anchor
User’s start point
Korean Palace Augmentation
MAR Content
PoI
http:.//Geovid.org
Meta-tagging
WHO WHEN WHERE WHAT HOW
+Virtual
Palace.{Own
er, Creator}
+Virtual Palace.{Created Time,
Modified Time, Accessed Time,
Duration}
+Virtual Palace Info.{Begin
Time, End Time}
+Virtual
Palace
Information.{L
ocation}
+Virtual Palace
Information.{Type,
Description}
+Virtual Palace VO.{Type,
String, Data File}
+Virtual
Palace
Info.{Condi
tion state}
WHO WHEN WHERE WHAT HOW
+Palace
Anchor.{Ow
ner, Author}
+Palace
Anchor.{Authored
time, Modified Time}
+Palace Anchor.{PoI
Coord., Contents
Coord., Coord. Type}
+Palace Anchor.{ID,
Tag, PoI.ID, MAR
Contents.ID}
+Palace Anchor.{Scale,
Interaction, Animation,
Accessibility}
WHO WHEN WHERE WHAT HOW
+Palace
Recognition.
{Owner,
Creator}
+Palace
Recognition.{Create
d Time, Modified
Time}
+Palace PoI.{Location,
Front Direction}
+Palace
Recognition.{Personal
Device Location, PoI 3D
Position}
+Palace PoI.{ID, Name,
Tag, Rec. Data, MAR
Anchor.ID}
+Palace Recognition.{ID,
Data Set ID, Candidate Sub
Data}
+Palace
Recognition.{Sc
ale, Gyroscope,
Compass,
Intrinsic
Parameter}
2
1
3
http:.//Geovid.org
Implementation in KCTM
 As implementation of the MAR reference model with 360 video, we are developing the
MAR service in K-Culture Time Machine project that provide personalized MAR storytelling
of the historical site in Korea
Implementation in KCTM
 As implementation of the MAR reference model with 3D 360 video, we are developing the
MAR service in K-Culture Time Machine project that provide personalized MAR storytelling
of the historical site in Korea
Implementation in KCTM
 As implementation of the MAR reference model, we are developing the phone-based MAR
service in K-Culture Time Machine project that provide personalized MAR storytelling of
the historical site in Korea
5W1H Metadata	Schema
who MAR	Contents.Essential.Who
when MAR	Contents.Essential.When
where MAR	Contents.Essential.Where
what
MAR	Contents.Essential.What,	MAR	Contents.{Type,	Reference,	MAR	
Anchor.ID},	VO.{Type,	String,	Data	File},	Media	Asset.{Type,	Data	File}
how MAR	Contents.Essential.How
Anchor
MAR Entity:
Metadata
1
3 MAR Conte
nt: Metadat
a
5W1H Metadata	Schema
who MAR	Anchor.Essential.Who
when MAR	Anchor.Essential.When
where
MAR	Anchor.{Essential.Where,	PoI	Coordinate,	Contents	Coordinate,	
Coordinate	Type}
what MAR	Anchor.{Essential.What,	MAR	Entity.URI,	MAR	Contents.ID}
how
MAR	Anchor.{Essential.How,	Scale,	Interaction,	Animation,	
Accessibility}
MAR Anchor:
Metadata
2
Anchor
MAR 
Entity
Virtual 
Contents
Anchor
5W1H Metadata	Schema
who {PoI,	Recognition	Data,	MAR	Entity	Information}.Essential.Who
when
{PoI,	Recognition	Data,	MAR	Entity	Information}.Essential.When,	MAR	Entity	
Information.{Begin	Time,	End	Time}
where
{PoI,	Recognition	Data,	MAR	Entity	Information}.Essential.Where,	
PoI.{Location,	Front	Direction},	Recognition	Data.{PoI	3D	Position,	GPS	
Location},	MAR	Entity	Information.{Location}
what
{PoI,	Recognition	Data,	MAR	Entity	Information}.Essential.What,	
PoI.{Recognition	Data.ID,	MAR	Entity.URI},	Recognition	Data.{Data	Set	ID,	
Sub‐Data	Set	ID, Current	Sub	Data,	Candidate	Sub	Data},	MAR	Entity.{URI,	
Label,	MAR	Entity	Information.ID,	PoI.ID,	MAR	Anchor.ID,	Property.URI},	
MAR	Entity	Information.{Type,	Description},	Property.{URI,	Label},	Literal	
Value.{Type,	Value}
how
{PoI,	Recognition	Data,	MAR	Entity	Information}.Essential.How,	Recognition	
Data.{Scale},	MAR	Entity	Information.{Condition	state}
PoI
MAR Entity
Information
Related
MAR 
Entity
Property
5W1H Metadata	Schema
who ‐
when ‐
where GPS,	Gyroscope,	Compass
what ID
how Intrinsic	Parameter
Essential:
Metadata
1 2 User:
Metadata
5W1H Metadata	Schema
who Creator,	Owner
when Created	Time,	Modified	Time
where ‐
what ID,	Name,	Tag
how ‐
Essential
Metadata
Anchor
PoI
MAR 
Contents
MAR 
Anchor
Recognition 
Data
MAR Entity 
Information
MAR execution engine
Contextualized PoIs
Context Analyzer 
PoI Analyzer
User Localization
View Point Setting
PoI Analyzing
Context Processor
Context Integration
Context Inference
Context Management
Recognized
PoIs
User profile
User
Context
Sensors
Sensor
Data
Recognition
Data
PoI Recognition
Data Contextualized
PoIs
Contextualized
PoIsPoIs

Simulator 
(event, space)
Renderer
MAR Execution Engine
Contextualized PoIs
Context Analyzer 
External 
Services
Open API
Personal
Device 
Display
MAR 
Visualization
User
Command
MAR Contents
MAR 
Anchor
MAR 
Ontology
MAR
Entity
MAR
Entity 
Info.팔작
지붕
인정전_PoI1
Contents Retrieval
Click
MAR
Entity
Info
MAR
Contents

Simulator 
(event, space)
Renderer
MAR Execution Engine
Contextualized PoIs
Context Analyzer 
External 
Services
Open API
Personal
Device 
Display
MAR 
Visualization
User
Command
MAR Contents
MAR 
Anchor
MAR 
Ontology
MAR
Entity
MAR
Entity 
Info.이화
문양
인정전_PoI2
Contents Retrieval
Click
MAR
Entity
Info
MAR
Contents
Contents Retrieval

Simulator 
(event, space)
Renderer
MAR Execution Engine
Contextualized PoIs
Context Analyzer 
External 
Services
Open API
Personal
Device 
Display
MAR 
Visualization
User
Command
MAR Contents
MAR 
Anchor
MAR 
Ontology
MAR
Entity
MAR
Entity 
Info.
희정당
희정당_PoI1
Click
MAR
Entity
Info
MAR
Contents
Now, Metadata for MAR Services…
Is it useful or worthy of standard?
Finally, I prepared for NWIP…
MAR
Services
…
DB B
DB CDB D
DB E
“The future is already here. It is just not
uniformly distributed” -William Gibson (SF writer)
More Information
Woontack Woo, Ph.D.
Mail: wwoo@kaist.ac.kr
Web: http://uvr.kaist.ac.kr

170119 metadata representation_v2s