Keywords
ontology, sensor data, machine learning, knowledge discovery, environmental ontology learning
Start Date
1-7-2012 12:00 AM
Abstract
In environmental engineering, sensor measurement is a process undertaken with respect to an environmental domain, i.e. an area of interest, to measure a domain property. Knowledge of a domain, meaning its concepts and relations that hold among them, can be formally represented by means of ontology. Therefore, given an ontology for an environmental domain, it seems reasonable to suggest that sensor data acquisition can be translated into ontological knowledge acquisition. We demonstrate this translation for the domain of road vehicle classification by measurement of vibration. We show how supervised machine learning is applied to learn a function that maps sensor data to ontological concepts. Hence, we abstract from both the physical sensor layer and the sensor data layer by discarding raw measurement data and retaining the knowledge conveyed by these data. We show how rules can be used to infer new domain knowledge, such as vehicle velocity.
Making sense of sensor data using ontology: A discussion for road vehicle classification
In environmental engineering, sensor measurement is a process undertaken with respect to an environmental domain, i.e. an area of interest, to measure a domain property. Knowledge of a domain, meaning its concepts and relations that hold among them, can be formally represented by means of ontology. Therefore, given an ontology for an environmental domain, it seems reasonable to suggest that sensor data acquisition can be translated into ontological knowledge acquisition. We demonstrate this translation for the domain of road vehicle classification by measurement of vibration. We show how supervised machine learning is applied to learn a function that maps sensor data to ontological concepts. Hence, we abstract from both the physical sensor layer and the sensor data layer by discarding raw measurement data and retaining the knowledge conveyed by these data. We show how rules can be used to infer new domain knowledge, such as vehicle velocity.
