Five-Point Algorithm Solution with Initial value Estimation and Non-Linear Least Square Optimization

Abstract

This paper proposes five-point algorithm with initial value estimation and non-linear least square optimization. Five-point algorithm is a powerful method for relative orientation because it requires no initial assumption of camera position and attitude. This algorithm algebraically calculates an essential matrix from five point correspondences between two calibrated cameras. However, a result of five-point algorithm includes multiple solutions and some selecting process must be executed. The method proposed here tries to calculate the appropriate essential matrix with non-linear optimization with appropriate initial values, which is obtained with additional 3 pairs of corresponding points. In this paper the relationship between traditional parametric relative orientation and essential matrix is discussed, and after that five-point algorithm with non-linear least square optimization is introduced.