Graph-Based Topic-Focused Retrieval in Distributed Camera Network

Abstract

Abstract-Wide-area wireless camera networks are being increasingly deployed in many urban scenarios. The large amount of data generated from these cameras pose significant information processing challenges. In this work, we focus on representation, search and retrieval of moving objects in the scene, with emphasis on local camera node video analysis. We develop a graph model that captures the relationships among objects without the need to identify global trajectories. Specifically, two types of edges are defined in the graph: object edges linking the same object across the whole network and context edges linking different objects within a spatial-temporal proximity. We propose a manifold ranking method with a greedy diversification step to order the relevant items based on similarity as well as diversity within the database. Detailed experimental results using video data from a 10-camera network covering bike paths are presented.Index Terms-Distributed camera network, diverse and relevant ranking, graph-based modeling, information search and retrieval.

Context-aware graph modeling for object search and retrieval in a wide area camera network

“…Zhu et al [22] proposed a context-aware activity recognition and anomaly detection system leveraging spatial-temporal and scene contexts. This paper extends our earlier work [20] where we first introduced contextual links into a graph model. We propose a graph based methodology to represent relationships among camera observations.…”

“…Videos (640x480, about 20 frames per second with variable frame rate) are captured for several hours in an uncontrolled environment with complex shape and appearance changes in objects, wireless packet losses and irregular illumination variations. This is a challenging dataset for which most of the off-the-shelf object detection and tracking algorithms fail due to various reasons [20]. For all the application scenarios presented in this section, we set the confidence parameters α = 0.20, β = 0.40, γ = 0.40 and η = 1.…”

“…Tong et al [19] provide an efficient strategy to fuse information from multiple sources such as image level features and text based annotations. Similar to [15], [20], we pose the problem of user interaction as a unified ranking problem for identifying representative and diverse snapshots that match the user query.…”

Context-Aware Hypergraph Modeling for Re-identification and Summarization

“…Ni et al [2] introduced graph based models for object search and retrieval in camera networks. Xu et al [3] implicitly introduced contextual links into a graph model but they do not use appearance based information. Sunderrajan et al [4] explicitly modeled appearance, spatial-temporal and scene contexts into the graph model to improve the accuracy.…”

“…Let be the global affinity matrix that defines node and edge affinities such that and and otherwise (2) where encodes node-to-node similarity. For computing node-to-node similarity, color drift pattern is taken into account and it is given by (3) where and are LAB space color histograms extracted over superpixel regions.…”