An algorithm of pose estimation based on conic correspondences
Wenchao Wang, Xu Zhang, Yong Feng Qi, Wei Cheng · 2017
As we all know, in computer vision, conics are widely applied as one of the most underlying image features together with points or lines and sometimes the correspondences of the points or lines are not available, so under the circumstances, conics are also a good choice which could be identified robustly. This paper gives a linear algorithm of estimating pose of camera based on the correspondences of conic from world plane to image plane system. The numbers of pairs of conics needed are equal or greater to three. The core of the algorithm is the solution of homography. In this case, we give an overdetermined set of linear equations and the solution got by the singular value decomposition is the homography. Hence, the proposed formulation leads to an algorithm that is efficient and easy to implement. On the basis of homography is known, the pose of camera can be estimated with matrix operations. In addition, the validity and robustness of the algorithm is verified though experiments with simulation data and real measured data. The main novel points in this paper are: First, comparing with the traditional methods, we use conics with the advantages more common, easier to be extracted and more stable than points; Second, using matrix transformation and singular value decomposition only, avoids solving polynomials so that the complexity of the algorithm is reduced greatly.