ISWC2010 Evaluation

Ontology of Perception: A Semantic Web Approach to Enhance Machine Perception

Today, many sensor networks and their applications employ a brute force approach to collecting and analyzing sensor data, and ignore the semantics inherent in the environmental data. Such an approach often wastes valuable resources – including both energy and computational resources – by unnecessarily tasking sensors and generating observations of minimal use. People, on the other hand, have evolved sophisticated mechanisms to efficiently perceive their environment. Such mechanisms include the use of background knowledge to determine what aspects of the environment to focus and a strong interdependent relationship between our ability to observe and perceive. In this paper, we develop an ontology of perception – derived from cognitive theory – that may be used to more efficiently collect and analyze sensor data. We evaluate this approach by collecting and analyzing observations of weather conditions, and show up to 50% reduction in the number of observations necessary for analysis.

System Architecture

Perception Cycle

Figure 1. Perception Cycle.

Perception Process

Figure 2. Perception Process.

Observation Process

Figure 3. Observation Process.

Ontologies and Knowledge Bases

Used by Observation Process: Observation Knowledge Base
Used by Perception Process: Perception Ontology

Perception Datasets (generated during various executions of the perception cycle):

• 25 miles (17 observers)
• 50 miles (70 observers)
• 100 miles (170 observers)
• 200 miles (373 observers)
• 400 miles (516 observers)

Statistics

Percepts Generated during Evaluation: # and %

25 Miles

Figure 4.

50 Miles

Figure 5.

100 Miles

Figure 6.

200 Miles

Figure 7.

400 Miles

Figure 8.

Percepts Generated during Evaluation: Trends

Grounded Percepts

Figure 9.

Extraneous Percepts

Figure 10.