Context-Aware Graph Modeling for Object Search and Retrieval in a Wide Area Camera Network
This paper proposes a context-aware object search and retrieval in a wide area distributed camera network. With the proliferation of smart cameras in urban networks, it is a challenge to process this big data in an efficient manner. A novel graph based model is proposed to establish relationship between moving objects in the scene and provide a system to search and retrieve the object of interest. More importantly, we exploit the fact that the objects occurring in close proximity with respect to space and time are not completely independent and serve as context for each other. We provide a methodology for encoding several contextual information such as appearance, spatial-temporal, and scene contexts into the graph model to improve the overall accuracy. A manifold ranking strategy is used to order the items based on the similarity with an emphasis on diversity. Extensive experimental results on a ten camera network are presented.