|
Middleware for Large Scale Surveillance (MILS)
As shown in above figure, MILS provides the data management
services needed to build a
large scale smart surveillance application.
While
MILS builds on the extensive capabilities of IBM’s
Content Manager
and DB2 systems, it is essentially independent of these products and
can be
implemented on top of 3rd party relational databases. MILS provides the
following functionalities
- Converting the Viewable Video
Index (VVI) to Relational Tables: MILS is
capable of taking the VVI generated by the IBM:SSE and inserting the
data index
into a relational database, thus allowing for SQL based querying of the
index
data by an application.
- Query Services for
Surveillance Data: MILS
provides applications with a set of services for querying the
surveillance
data. The queries supported by MILS cover a wide range of commonly used
queries. The following are types of queries that will be supported by
MILS.
Each of these query types is illustrated by an example. The basic unit
of
storage in the database is a continuous movement event, for example, a
car
entering the camera field of view at time T1 and moving continuously to
leave
the field of view at T2, would be represented as a basic unit called an
activity.
- Time Queries: Show all activities between (7 PM Apr 17 2004) and (9
PM Apr 17 2004).
- Object Size Queries: Show all activities where the object
size O is (S1 < O < S2).
- Object Class Queries: Show all
activities with objects
of class C1. Current set of classes are
limited to Vehicles, Single Person and Group.
- Object Motion Queries: Show all
activities where the
object moved with speed V where, (V1 < V < V2) or where an object
moved
in a direction D (specified as an angle in the image plane).
- Context-based object content
similarity Queries: Show
all activities where there were “blue cars” or “cars similar to this
car –here
the user specifies an example car through an image”.
- Queries based on region/location of
interest: Show all
activities that occurred within a specified polygonal region.
- Combination Query: A query which
combines the one or
more of the above specified query types.
- Interactive object/event
classification based on
relevance feedback
- Data Mining Services: MILS will
support the following
data mining applications.
- Site activity summarization
- Site activity pattern discovery
- Automatic abnormal
event/object detections
Other Research Areas:
|
Other
projects in