A Theoretical Framework for Activity Classi fication

Hamid Krim
North Carolina State University

Shape analysis is playing an increasingly important role in many applications where object classifi cation and understanding are of interest. Solutions to many existing as well as new emerging applied problems (e.g, object recognition, biometrics etc.) crucially depend on
object modeling and their parsimonious representation.
Modeling an active silhouette in a video sequence provides a good solution for activity surveillance. We pose this problem as one of tracking a ow of shapes as entities on a curved space. We first propose a stochastic model for a
ow on a manifold to carry out classifi cation of diff erent processes. We then exploit this insight to develop a tracking filter of these shapes and subsequently propose a generative model useful in a variety of applications. We subsequently propose a generative model for human activity. We provide substantiating illustrations.

Back to Large Scale Multimedia Search