MULTIMEDIA DATABASES AND QUERYING TECHNIQUESBy: Rohit Kulkarni
CS 2310 – Spring 2008
AGENDA
Background
Problem Definition
Challenges
BACKGROUND
What is MMDBMS?
Normalization framework
What is data fusion?
The general problem
Querying technique
MMDBMS ARCHITECTURE
REQUIREMENTS FOR MMDBMS
Traditional DBMS capabilities Huge capacity for storage management Information retrieval capabilities Media integration, composition and
representation Multimedia query support Multimedia interface and interactivity Performance
ISSUES IN MMDBMS
Multimedia data modeling Multimedia object storage Multimedia integration, presentation and
QOS Multimedia indexing, retrieval and browsing Multimedia query support Distributed multimedia database
management System support
PROBLEM DEFINITIONS
Extended Dependencies
The relational model
Similarity theory
Tuple distance function
AN EXAMPLE
Define a functional dependency between attributes FINGERPRINT and PHOTO of police database, and use the fingerprint matching function FINGERCODE for comparing digital fingerprint [JPH00], and the similarity technique used by QBIC for comparing photo images, we would write as follows
FINGERPRINTFINGERCODE(t’) PHOTOQBIC(t’’)
INFERENCE RULES FOR MFDS
Reflexive rule Augmentation rule Transitive rule Decomposition rule Union rule Pseudotransitive rule
NORMAL FORMS IN MULTIMEDIA DATABASES
Normal forms are used to derive database schemes that prevent manipulation anomalies
Similar anomalies can arise in multimedia database
Types of normal forms are1MNF, 2MNF , 3MNF and 4MNF
SENSOR DATA FUSION
Background:
Need for Multiple Sensors:- data from a single sensor yields poor results in object recognition
Sensor Management Model
SENSOR MANAGEMENT MODEL
SENSOR DATA FUSION
Problems:
Association of objects from different Sensors
Tracking
NEED FOR QUERY TECHNIQUE
• Problem with existing query techniques
• Why not SQL?
• To support the retrieval and fusion of multimedia information from multiple sources and distributed databases, a spatial/temporal query language called QL has been proposed
FEATURES OF QL
Easy to learn as syntax is similar to SQL
Allows user to specify queries for both Multimedia data sources and Multimedia databases
Supports multiple sensor sources and systematic modification of queries
OPERATOR CLASSES
The operators in QL can be categorized with respect to their functionality.
The two main classes are: transformational operators (the σ‑operators) fusion operators (the ‑operators).
TRANSFORMATIONAL OPERATORS
Definition: A σ‑operator is defined as an operator to be applied to any multi-dimensional source of objects in a specified set of intervals along a dimension. The operator projects the source along that dimension to extract clusters
TRANSFORMATIONAL OPERATORS (CONTD)
As an example, if we write a σ‑expression for extracting the video frame sequences in the time intervals [t1-t2] and [t3-t4] from a video source VideoR.
The expression will be
is σtime([t1-t2], [t3-t4]) VideoR
where VideoR is projected along the time dimension to extract clusters (frames in this case) whose projected positions along the time dimension are in the specified intervals.
FUSION OPERATORS
Much more complex as it deals with Sensor data fusion
Requires input data in different time periods from multiple sensors
The output of the fusion‑operator is some kind of high level, qualitative representation of the fused object, and may include object type, attribute values and status values.
IS THERE A MOVING VEHICLE PRESENT IN THE GIVEN AREA AND IN THE GIVEN TIME INTERVAL?
IS THERE A MOVING VEHICLE PRESENT IN THE GIVEN AREA AND IN THE GIVEN TIME INTERVAL?
• Corresponding query:type,position, direction
(motion(moving) type(vehicle)
xy(*)
(T)T mod 10 = 0 and T>t1 and T <t2
media_sources (video)media_sources
ype (vehicle) xyz(*)
(T) T>t1 and T<t2
media_sources(laser_radar) media_sources)
EXPERIMENTAL PROTOTYPE
CHALLENGES
Handle large number of different sensors
Replacing manual query with a semi-automatic or fully automatic query refinement process
APPLICATIONS OF MMDBMS
Education- digital libraries, training, presentation
Healthcare- telemedicine, health information management
Entertainment- interactive TV, video on demand
Information dissemination- news, TV broadcasting
And many more!
REFERENCES
Intelligent Querying Techniques for Sensor Data Fusion by Shi-Kuo Chang, Gennaro Costagliola, Erland Jungert and Karin Camara
A Normalization Framework for Multimedia Databases by S.K. CHANG, V. DEUFEMIA, G. POLESE
Querying distributed Multimedia databases and data sources for sensor data fusion by S.K.Chang, Gennaro Costagliola, Erland Jungert and Francesco Orciuoli
Multimedia database management-requirements and issues by Donald A. Adjeroh and Kingsley C. Nwosu
Fuzzy Queries in Multimedia database system by Ronald Fagin Bayesian Approaches to Multi-Sensor data fusion by Olena
Punska, St. John’s CollegeMultimedia database management-requirements and issues by Donald A. Adjeroh and Kingsley C. Nwosu Querying distributed Multimedia databases and data sources for sensor data fusion by S.K.Chang, Gennaro Costagliola, Erland Jungert and Francesco Orciuoli
MULTIMEDIA DATABASE AND QUERYING TECHNIQUES
By: Rohit KulkarniCS 2310 – Spring 2008
Thank you