Event Detection and Fusion Model for Semantic Interpretation of Monitored Scenarios within ASIMS Architecture - Foundations on Natural and Artificial Computation - page 523

Fig. 2. Event Detection, Fusion and Activity Detection

The Common Data Model stores sensor information and the results from
executing each of the algorithms: segmented images, number of objects in the
scene, global position, speed, direction, shape, colour, class to which they belong,
properties, detected events, etc. There is also a time marker or timestamp to
label the information captured. All the modules have reading access, but they
only allow writing in those data that they generate as output. The Central Node
has two differentiated stages (Figure 2): Event Fusion and Activity Detection.
The first of these is in charge of filtering information from all the Local Agents.
The second detects global activities and generates the necessary “warnings” in
pre-defined situations requiring human decisions.

3E entD on

In the description of the architecture in the previous chapter, Event Detection
is in the last level of the Remote Nodes. The aim of this module is to identify
patterns existing in the data structure of the Common Data Model, described
in the previous chapter, in order to detect patterns that are produced in the
monitored data. The characterisation of these patterns (that generate events)
can be represented with finite automata for their resolution. Each of the events
to be detected has its own associated finite automaton containing the necessary
states to detect an event with its transition state. Thus, if we have three events
to identify and two objects in the scene, three states will be generated by each
Object:

Object 1[ Event 1= q 0; Event 2= q 0; Event 3= q 0]

Object 2[ Event 1= q 0; Event 2= q 0; Event 3= q 0]