Distributed camera network localization

Abstract

Localization, estimating the positions and orientations of a set of cameras, is a critical first step in camera-based sensor network applications such as geometric estimation, scene reconstruction, and motion tracking. We propose a new distributed localization algorithm for networks of cameras with sparse overlapping view structure that is energy efficient and copes well with networking dynamics. The distributed nature of the localization computations can result in order-ofmagnitude savings in communication energy over centralized approaches. I. INTRODUCTION Increasingly powerful microprocessors, along with recent advances in micro-sensor technology, wireless communications, and power efficiency, provide new opportunities to sense, compute, and discover via wireless sensor networks. A sensor network consists of a collection of nodes placed or scattered throughout an environment of interest. Each node consists of a power supply, set of sensors, data processor, and a wireless communication system. In this paper, we focus on networks of digital cameras. As CCD and CMOS image/video cameras become smaller, less expensive, and more power efficient, camera networks will become increasingly practical and widespread. Example camera network applications include most of the key tasks from computer vision, including low-dimensional correspondence searches, mosaicking through image rectification, view morphing, image-based navigation, and scene geometry extraction. Due to the high-dimensionality of image and video data, it is paramount in a camera network to distribute as much processing as possible locally within the network. A critical first step in any camera network deployment is localization; the camera nodes must determine their positions and orientations in three-dimensional (3-D) space. Most current sensor network localization techniques use acoustic delays and radio frequency (RF) intensities to create a set of pairwise distances in order to localize [1]. In contrast, little work has been done on distributed algorithms for camera localization via the images acquired at each node. In this paper, we develop the Distributed Alternating Localization-Triangulation algorithm (DALT), an iterative, distributed algorithm for camera network localization. This

3-D Target-Based Distributed Smart Camera Network Localization

“…Mantzel et al extract feature points by analyzing tracked motion, and correlate the features across views using time synchronization [1]. Mantzel et al compensate for inaccurate correlations by determining a subset that produces the essential matrix estimate with the least error according to the epipolar constraint.…”

“…The most recent solutions opportunistically search for robustly identifiable world features and correlate them between pairs of cameras with view overlaps [1], [2], [3]. Correlated features are used to estimate either the essential or fundamental matrix for two view overlapping cameras and which decomposed provides the camera pair's relative position and orientation, which is the data needed for network localization [4], [5].…”

Distributed computer vision algorithms through distributed averaging

“…We are not aware of any previous distributed version of GPCA [10]. Triangulation of 3-D points in a camera network has been studied in the context of camera localization [4,8]. In those works, however, the triangulation is performed independently at each camera, while our approach uses all the images simultaneously.…”

Wireless sensor node localization

“…Cameras with overlapping views can observe a set of feature points and deduce their relative positions. However, the obtained coordinates are relative to a scaling factor [7] that needs to be determined via other means. In simultaneous localization and tracking, a moving object is observed and tracked instead of a set of feature points [8].…”