Camera Network Calibration and Synchronization from Silhouettes in Archived Video

Abstract

In this paper we present an automatic method for calibrating a network of cameras that works by analyzing only the motion of silhouettes in the multiple video streams. This is particularly useful for automatic reconstruction of a dynamic event using a camera network in a situation where pre-calibration of the cameras is impractical or even impossible. The key contribution of this work is a RANSAC-based algorithm that simultaneously computes the epipolar geometry and synchronization of a pair of cameras only from the motion of silhouettes in video.Our approach involves first independently computing the fundamental matrix and synchronization for multiple pairs of cameras in the network. In the next stage the calibration and synchronization for the complete network is recovered from the pairwise information. Finally, a visual-hull algorithm is used to reconstruct the shape of the dynamic object from its silhouettes in video. For unsynchronized video streams with sub-frame temporal offsets, we interpolate silhouettes between successive frames to get more accurate visual hulls. We show the effectiveness of our method by remotely calibrating several different indoor camera networks from archived video streams.

Camera Alignment Using Trajectory Intersections in Unsynchronized Videos

“…They synchronize views by rank constraints on matrices that capture either the linear combination between points in two views or the brightness measurements of image patches. Sinha and Pollefeys [17] simultaneously calibrate and synchronize cameras in a network, but require the silhouette of a person instead of the trajectory alone.…”

On Cross-Spectral Stereo Matching using Dense Gradient Features

“…[20] and [14] assume that an estimate of the projective calibration is available (or can be computed from the static background), and utilise the epipolar constraint and the tri-focal transfer error, respectively, to estimate the synchronisation parameters. [25], on the other hand, can work with fully dynamic scenes, such as a blue screen scenario, and given an accurate foreground segmentation, computes a joint calibration and an index shift estimate.…”

“…[26], [20], [14] and [25] are unique in the sense that they discuss scenarios with more than 2 cameras. However, [25] and [14] assume a constant frame rate, and estimate only the temporal offset.…”

“…However, [25] and [14] assume a constant frame rate, and estimate only the temporal offset. [20] returns a frame rate and an offset for each camera in the network, but only if there are points visible in all frames, an assumption which imposes restrictions on the content and the camera placement.…”

“…Unlike the sequential method presented in [14], a graph is not order dependent, and can represent the relationships between non-neighbouring pairs. Our approach is superior to the graph-based method of [25], as it can incorporate more constraints, in the form observed index correspondences. Also, unlike [14] and [25], it does not assume a known frame rate, and unlike [20], existence of points visible in all cameras.…”