Identification of alarming behaviour introduced by monitoring based in the integration of ontologies

Abstract

This article presents a system of monitoring of human activities based in the open possibilities by the integration of different ontologies and semantic buildings, in an alone ontology called OSU (Ontology for Surveillance and Ubiquitous Computing), for this; it uses the definition and composition of events, from the ones of minor level (events that interpret changes of state in the sensors), until events interpreting high level activities as interpreted by an expert, particularly significant or alarming. To validate the process, we worked in two stages and with two target-activities (alarming behaviour): During monitoring at home through sensors of movement situated in determined positions where can be done a follow-up of the evolutions of the person that lives in this place; here the target-activity (alarming behaviour) to be identified is night_wandering. The other scenario happens in one of the supermarket where is carries out a follow-up of a person, by means of video, consequently identifying the target-activity marauder. The results of the validation are encouraging; demonstrating that OSU is valid for the modelling of generating events alarms or can be used in future work for modeling �other generate events alarms� based on our proposal.