Automatic Learning of Typical Paths in Video Surveillance
Jorge Vicente, Nunes, Luís · Repositório do ISCTE-IUL · 2013
This paper presents an approach to the acquisition of typical paths tread by people in large environments, and signal deviations. This work is part of a larger project, tested both in simulations as well as using real footage. The adaptation required for the continuous update of typical paths is achieved using Self-Organizing Maps (SOM), supplemented with original solutions for map-selection.