Classificação de grupos utilizando informações de geometria e detecção de atividades intragrupo

Abstract

The main goal of this work is to propose a model for an automatic (our semiautomatic) classification of groups using geometrical properties and for detecting intragroup

activity based on video sequences. For group classification, a tracking algorithm is applied to obtain the position of each person across time, and the relationships among these people and their orientation are used to detect and classify groups based on sociological information (proxemics, interpersonal distance, etc). The geometry of the group, as well as its temporal evolution, are used to provide additional information on the group. To detect intragroup activity, the temporal evolution of blob áreas related to tracked people is explored.

Regarding possible applications of this work, an example could be the detection and automatic classification of small groups in a shopping center, in order to extract behavior pattern according to this study