Abstract:

Abstract: Multiple camera systems (MCS) have attracted increasing attention in machine vision. Regarding pose estimation of MCS, a novel integrated reconstruction and orthogonal iteration algorithm is proposed. A general camera model is introduced to formulate MCS to derive the algorithm. To make full use of  the redundant measurements of MCS, the observed reference points are classified, and a weighted objective function is constructed. Using the objective function, advantages of the optimal absolute orientation solution (OAOS) and the orthogonal iteration (OI) algorithm are combined. Introduction of 3D reconstructed information not only overcomes the pose ambiguity of OI but also improves convergence and robustness in pose estimation. The effects of reconstructed error on final estimation are reduced using an adaptive weighting method. Simulation results show effectiveness and superiority of the proposed algorithm.

Key words: machine vision, multiple camera systems, pose estimation, pose ambiguity, orthogonal iteration, adaptive weighting