Object tracking using wireless camera networks

Henry P Medeiros, Purdue University

Abstract

This dissertation presents all the components of a complete system to perform collaborative object tracking with wireless camera networks (WCNs). Whereas the design challenges of WCNs are similar to those of wireless sensor networks (WSNs), most methods devised for WSNs cannot be directly applied to networks of wireless cameras. The dynamic clustering protocols (for collaborative processing and data aggregation) devised for WSNs do not consider the fact that physical proximity between cameras does not imply common sensing regions. In addition, aggregation of the data generated by the camera nodes for moving targets presents new challenges, such as the need to estimate in real-time the constantly-changing target state on the basis of the visual information acquired by the nodes at different time instants. To address these and other issues unique to wireless camera networks, this dissertation presents a distributed object tracking system for WCNs that employs a cluster-based Kalman filter. Our method requires the transmission of significantly fewer messages than a centralized tracker that naively transmits all the local measurements to the base station. It is also more accurate than a decentralized tracker that employs linear interpolation for local data aggregation. For the new generation of smart cameras that employ sophisticated computing architectures, this dissertation also includes a histogram-based particle filter for object tracking on smart cameras based on single instruction multiple data (SIMD) parallel processors. This algorithm performs robust object tracking on a low-power SIMD processor at speeds comparable to those obtained with desktop computers.

Degree

Ph.D.

Advisors

Kak, Purdue University.

Subject Area

Electrical engineering

COinS