An Algorithm for cameras networks calibration
Abstract:
In this talk we will deal with the problem of the angular calibration for a network of cameras, namely the problem of estimating a common orientation reference frame. In the proposed set–up each camera obtains noisy measurements of its relative orientation with respect to some other cameras. The set of measurements can be described by a graph having the cameras as nodes and having an edge between two cameras if a relative orientation measurement is available. We propose a novel two-step algorithm based on a choice of a basis for the set of graph cycles. The first step consists in computing a set of integer numbers, which provides a first rough estimate of the orientations. The second step exploits this information to build up a suitable quadratic minimization problem. Two actual implementations, corresponding to two different basis of cycles, are described and compared in terms of the worst–case scenario. Finally, through numerical simulations the algorithm is compared with another algorithm proposed in the literature for solving the same problem.