Event Detection and Fusion Model for Semantic Interpretation of Monitored Scenarios within ASIMS Architecture - Foundations on Natural and Artificial Computation - page 526

problem names have to be standardised for the whole system. Using a method
based on the global positioning of identified objects, equivalence tables are estab-
lished (Table 1) in accordance with the Remote Node sending this information.
With these tables and a translation service for names, the names for each scene
object are standardised.
Fus i on: In centralised systems with different clients sending information, con-
tradictory or redundant information can be received at the same instant of time.
This stage focuses on solving redundancies and contradictions from the different
Remote Nodes. For this a model of the scenario is used where the information
on the positions of the Remote Nodes and all the scenario objects is. To iden-
tify and solve redundancies the Events are grouped with the same timestamp
according to their spatial reference in order to facilitate the process. Once the
classification has been done a search is done for equivalent events. This equiva-
lence is established in accordance with the Remote Sensor Nodes and the event
equivalence formulation.
After the fusion, the resulting events are transmitted to the activity detection
system (next processing level, which composes-abstracts events with greater se-
mantic content) like that described in the article by R. Martinez-Tomas & A.
Rivas-Casado [11]. This system is in charge of identifying spatial-temporal ac-
tivities in the controlled environment in order to provide semantic information
to the sequences of events that are perceived in the capture process.
The figure 5 represents a possible fusion situation where three sensors are
monitoring a common area. Each Remote Node Event Detection Module iden-
tifies the event “At” at the same instant of time. The structure of the event “At
consists of the set of attributes: human label ?h, position x ?x, position ?y and
the instant of time ?t when the event is generated:

At (( human ? h )( x ? x )( y ? y )( time ? t ))

In the acquisition stage, all the names of the objects are extracted and compared
in the table of existing names. Each remote node and name is assigned its global
object name. This is done by positioning each object in the global world. In our
example the acquisition stage receives three events “At” from three remote nodes
(Node 1, Node 2 and Node 3) (Table 1)

Tabl e 1 . Name equivalence table

Remote Node Object name GlobalObject

Node 1 Human59 GlobalHuman65

Node 2 Human48 GlobalHuman65

Node 3 Human73 GlobalHuman65

The fusion of two or more events “At” generates a new event “At with the
standardised name of the object, the mean of the positions x of all the events
“At”, the mean of the positions and all the events “At” and the timestamp. The
table 2 shows the group of events from the three nodes and how they merge to
give a central event.