Asynchronous Distributed Calibration of Camera Networks

This work focuses on the problem of calibrating planar networks of cameras in a distributed fashion. The camera network is modeled by a graph, and along each edge a noisy relative angular measurement is available. The goal is to achieve the absolute orientation of each camera with respect to a fixed external reference frame, in order to be able to perform monitoring and patrolling tasks. The idea is to exploit the cycles in the graph, along which all relative measurements sum to zero, in order to eliminate the noise. We design a distributed algorithm for the cameras to autonomously calibrate and we adopt an asynchronous gossip-like communication protocol. The proposed algorithm is proved to converge, almost surely and in the mean square sense, to the set of angles with zero cycle error. Finally, numerical experiments are presented to compare the performance of the algorithm on different graph topologies.