PhD forum: Hierarchical feature scheme for object recognition in visual sensor networks

Vildana Sulic, Janez Pers, Matej Kristan, Stanislav Kovacic

Abstract: We propose a framework for hierarchical feature distribution scheme for object recognition in a network of visual sensors. The key of our approach is the principle that individual nodes in the network hold only a small amount of information about objects seen by the network, however, this small amount is sufficient to efficiently route queries through the network. Preliminary results suggest that amount of data transmitted through the network can be reduced in comparison to simpler feature distribution schemes.