Modelli 3D del corpo umano per la riacquisizione di persone

Abstract: People Re-identification is a fundamental task for the analysis of long-term activities and behaviors of specific people. It aims at associating multiple instances of the same person acquired from different points of view or after a temporal gap. It can be very useful for people analysis in security, as a data association method for long term tracking in surveillance, as a soft-biometry task or as an image retrieval tool on video archives in forensics and other security-related applications. In this thesis the re-identification task is thoroughly analyzed in all it's aspects, limits and requirements, describing and evaluating the methods available in the literature. A new line of research is proposed: the exploitation of non-articulated and articulated 3D body models for pedestrian re-identification. Capabilities ad limits of this new paradigms are discussed and demonstrated through a comprehensive experimental evaluation.

Citation:

Baltieri, Davide "Modelli 3D del corpo umano per la riacquisizione di persone" 2013

 not available

Paper download:

• Author version: