A Robust O(n) Solution to the Perspective-n-Point Problem

We propose a noniterative solution for the Perspective-n-Point (Pn P) problem, which can robustly retrieve the optimum by solving a seventh order polynomial. The central idea consists of three steps: 1) to divide the reference points into 3-point subsets in order to achieve a series of fourth order polynomials, 2) to compute the sum of the square of the polynomials so as to form a cost function, and 3) to find the roots of the derivative of the cost function in order to determine the optimum. The advantages of the proposed method are as follows: First, it can stably deal with the planar case, ordinary 3D case, and quasi-singular case, and it is as accurate as the state-of-the-art iterative algorithms with much less computational time. Second, it is the first noniterative Pn P solution that can achieve more accurate results than the iterative algorithms when no redundant reference points can be used (n 5). Third, large-size point sets can be handled efficiently because its computational complexity is O (n).