Error analysis of the motion estimation based on SVD theory
Wei Shao, Lin Meng, Hao Hua Qin, Hai Yan Chen, Shu Lin Sui · 2014
For the two graphics whose internal parameter matrix is known, their epipolar geometry can be determined through the image coordinate matching points which have been observed. However, matching noise will have some impact on the estimation of motion parameter. In this paper, the essential matrix is decomposed by factorization algorithm to get external parameters, which is based on the Singular Value Decomposition (SVD) theory. The error propagation of motion estimation is analyzed based on matrix perturbation and covariance propagation theory of linear model. Meanwhile, the expression of motion estimation error which changes with the essential matrix error is deduced. Finally, the correctness of the deduced results is proved by matlab simulation experiment.