Efficient Fundamental Matrix Estimation for Robotic Visual Servoing Based on Continuous-time Optimization

This paper deals with the problem of estimating the fundamental matrix in real time for robotic visual servoing. In order to improve the efficiency of the fundamental matrix estimation, a new method incorporating the rank-2 constraint directly is proposed based on continuous-time optimization. By designing a proper projection matrix, the estimation is transformed into an integration process of a continuous-time dynamical equation. Simulation and real experiments show that, compared to conventional discrete-time optimization methods, the proposed method could achieve similar accuracy while converging faster (more than 7 times faster when the number of point correspondences is 1000).