PhD Defense: "PhD Defense: Object Tracking and Searching in Distributed Camera Networks"

Zefeng Ni

December 1st (Thursday), 3:00pm
Harold Frank Hall (HFH), Rm 4164

Technological advances have created new opportunities and challenges towards distributed camera networks. One of the main challenge is how to effectively utilize network-wide knowledge under the constraint of distributed processing. This dissertation addresses the challenge and proposes novel distributed approaches for two basic vision tasks in a camera network: 1) distributed object tracking and 2) browsing and searching objects.

In this dissertation, object tracking is formulated as a global Bayesian estimation problem that is realized through a distributed Monte-Carlo sampling implementation. At each camera node, a local particle filter tracker with discriminative appearance model is used for image plane tracking. Two novel distributed active fusion schemes are proposed to facilitate the collaboration among the cameras for joint tracking. An on-line learned discriminative model is used for enforcing appearance consistency by weighting particles from the local visual tracker. The object’s ground plane motion consistency is enforced by correcting the local visual tracker’s particles that deviate from the ground plane estimate based on the information shared across the views. The proposed method allows an efficient closed loop interaction between object’s local tracking module and the global fusion schemes for a robust joint tracking. Experiment results verify and quantify the efficacy of the proposed methodology for human tracking in a camera network.

The second contribution of this dissertation is a novel system to assist human image analysts to effectively browse and search for objects in a camera network. In contrast to the existing approaches that focus on finding global trajectories across cameras, the proposed approach directly models the relationship among raw camera observations. A graph model is proposed to represent detected/tracked objects, their appearance and spatial-temporal relationships. In order to minimize communication requirements, raw video is processed at camera nodes independently to compute object identities and trajectories at video rate. However, this would result in unreliable object locations and/or trajectories. The proposed graph structure captures the uncertainty in these camera observations by effectively modeling their global relationships, and enables a human analyst to query, browse and search the data collected from the camera network. A novel graph ranking framework is proposed for the search and retrieval task, and the absorbing random walk algorithm is adapted to retrieve a representative and diverse set of video frames from the cameras in response to a user query. Preliminary results on a wide area camera network are presented.

About Zefeng Ni:

Zefeng Ni received his bachelor's degree in Computer Engineering from the Nanyang Technological University (NTU), Singapore in 2004. He worked as Project Officer at NTU from 2004 to 2006. He received the M.S degree in Electrical Engineering from the University of California Santa Barbara in 2007. Zefeng is currently pursuing a Ph. D. degree in Electrical Engineering under the guidance of Prof. B. S. Manjunath. His research interests lie in multi-camera video processing and analysis.

Hosted by: Professor B. S. Manjunath