Robust structure from motion approach based on multi cameras

Considering the effects of large-scale noises on measurement matrix, this paper proposes a robust framework to realize structure from motion based on multi-cameras. By introducing a mathematical model for recovering the low rank matrix, the problem of incomplete measurement matrix solution caused by missing data and large-scale noises is converted into that of a convex optimization, which can be solved using robust principal component analysis(RPCA) method; and the matrix factorization theory is used to achieve the structure from motion with multi-cameras. Experiment results show that the proposed framework can effectively solve the problem of structure from motion with multi-cameras caused by the existence of large scale noises, and accurately reconstruct the three-dimensional structure.