Multimedia target tracking through feature detection and database retrieval

The real-time detection and tracking of moving objects is a challenging task and automatic tools to identify and follow them are often subjected to constraints regarding the environment under investigation or the full visibility of the targeted object. Exploiting the possibility of a multi-source acquisition in the targeted scene, firstly detection is performed by means of characteristic features extraction and storing in a database; secondly, the tracking task is approached using algorithms, where automatic search involves occluded or masked targets in the scene. This latter problem is solved through database retrieval, based on well-defined multi-modal features. The method has been tested on case studies regarding the identification and tracking of animals moving at night in an open environment (i.e. natural reserves or parks), and the surveillance of known scenes for unauthorized access control.