Information Technology Reference
In-Depth Information
Event Detection and Fusion Model for Semantic
Interpretation of Monitored Scenarios within
ASIMS Architecture
Ángel Rivas-Casado and Rafael Martínez-Tomás
Dpto. Inteligencia Artificial. ETSI Informática, Universidad Nacional de Educación a
Distancia, Juan del Rosal 16, 28040 Madrid, Spain
{arivas,rmtomas}@dia.uned.es
Abstract. Semantic interpretation of monitored scenes implies the well-
known problem of linking physical signals received by sensors with their
meaning for a human. Our line of work aims to develop a global ar-
chitecture, which we call “Architecture for Semantic Interpretation of
Monitored Scenarios (ASIMS)”, to integrate all the information pro-
cessing abstraction levels, from the sensory agent process and coherent
fusion of agent results to behaviour and situation identification. This
work presents a specific structure for the acquisition and fusion of events
from the object level, which forms part of the global ASIMS structure.
In remote processing nodes, events caused by variations in magnitudes
in the common data model are identified via finite automata models.
These events are merged by the central node to solve the usual problems
of centralised systems: synchronisation, redundancy, contradiction and
heterogeneity of the information that they receive from different sources.
For this we have broken down the fusion mechanism into three stages:
Synchronisation, Standardisation and Fusion, which are described in the
article with simple examples.
1
Introduction
The overall problem that we wish to address is that of interpreting the activities
in the scene being monitored. Of course, it is the old problem of associating sensor
signals with semantics, with a description similar to human comprehension, like
a human understands or identifies the scene. Possibly the most generic term is
that of “interpretation”: a message is interpreted, a cardiogram is interpreted,
a video is interpreted, etc. We specifically want to interpret what happens in a
monitored scenario, identify events and activities. For this sensors are situated
and their signals are processed in the hope of finding spatial and time data
patterns or associations. These associations make it possible to ascend step by
step in successive abstraction levels to reach conclusions connected with decision-
taking: a situation has been identified that requires a coherent response, either
by a security guard or doctor, an alarm centre, an automated emergency system,
etc.
Search WWH ::
Custom Search