A Multisensory Monitoring and Interpretation Framework Based on the Model–View–Controller Paradigm - Foundations on Natural and Artificial Computation - page 447

Table 1. Event structure

<events>
<event name=Run>
<object>1</object>
<speed>5</speed>
<probability>0.7</probability>
<timeStamp>1256953732</timeStamp>
</event>
<event name=Walk>
<object>2</object>
<speed>1</speed>
<probability>0.6</probability>
<timeStamp>1256953735</timeStamp>
</event>
</events>

Table 2. Main features of the common model

Level Parameters Details

Acquisition - Acquired image list

- Data from sensors ⇒ XML
-IdTime

Data collection
Information sources management
Synchronization

Sensor Fusion - Fused image list Create new images from the acquired ones

Localization & Filtering - Localization & Filtering im-

age list

Isolate spots containing the objets of interest

Localization & Filtering
Fus ion

- Fused image list Merge localized & filtered images

Blob Detection Blob list containing:

-IdTime
-Contours
- Box (X,Y,Z,W,H,D)
- Color and brightness

From pixel-level information to blob-level
Obtain parameters to characterize the spots (coordinates,
color info, contours, etc.)

Object Identification Object list containing:

-IdTime
- Box (X,Y,Z,W,H,D)
- Real position
-Label
-Contours
- Characteristic Points
- Direction
- Speed

Blob-level information to object level
Information remain among iteration, being updated
Calculation of parameters defining object motion
Mapping of the image coordinates of the objects into the
real world

Object Classification -Class

- Orientation

Update object list
Provide information concerning “what” the objects are and
their orientation

Object Tracking - Trajectory Calculation of object trajectories

Match objects changing the sensors' detection field

Event Detection - Event list Instantaneous event detection (semantic primitives)

Events are wrapped into XML structures

Event Fusion - Fused event list Performed by central node

Update event list, discarding mismatching events and unify-
ing repeated ones
Synchronization is a key aspect (when to fuse)
Feedback to local nodes according to their received events

Activity Detection - Activity list Global view of the scenario

Spatial and temporal information

Scenario Modeling Scenario map

- Light conditions
- Temperature
- Sensors and their ranges
- Detection distance

Related sensors information
Global and local modeling

Event Fusion: This event fusion level allows the consideration of new infor-
mation sources (the remote nodes), even though, at this level, sources provide
events. Event fusion provides a general view of the scenario by merging all events,
and eliminating wrong events. This again brings up the necessity for a good syn-
chronization of the nodes. The operation of this level generates an event structure
(as shown in Table 1) used as input by the next level.