Abstract :
The characterizing of persons in real time is a topic of interest in computer vision. In this work we develop an method of persons characterization. Particularly, the proposal is divided in 3 stages: 1) In a first stage, Mixture of Gaussians (MOG) is used to detected each objection movement,the contour of the foreground isanalysed with the radius distribution. Finally, for each block a measure of dispersion based in radius distribution, considering the maximum sparse criterion is proposed.
Keyword :
persons, characterization, method, tracking