Aplicación de tecnologías semánticas y técnicas de reconocimientos de objetos para la identificación de armas de fuego en video

Abstract

The implementation of the Digital Terrestrial Television in Ecuador has opened a wide field of action for projects and investigations related to this topic. From this context, the analysis and classification of televisual violent content becomes a fundamental task to accomplish in order to achieve a proper television programmes presentation to the audience. This project proposes the development of a process for the semantic annotation of firearms recognized in videos, with the purpose of generating a semantic repository containing the collected information so that it could be queried afterwards. To do this, a first prototype was developed which allows the detection and recognition of firearms appearing in video. Furthermore, an ontology was generated to model firearms and its classification by using NeOn methodology. This ontology was later fused with another ontology which models televisual content. Subsequently, a second prototype was developed to take the recognized firearms data, semantically annotate it using the generated ontology and later store it in a repository. Finally, a third prototype allowed the users to query the semantic repository. With the aim of contributing the addressed area of knowledge, a consolidated process is proposed. This process includes the whole executed work and serves as a guide for other projects implying objects recognition and the application of semantic technologies for their annotation.