Robust Fitting Method and Application Based on Geometric Invariance
Abstract：Fitting curves or surfaces is a fundamental problem, which has plentiful applications in computer vision, computer graphics, robotics and et al. Due to noise and occlusions, this problem is not simple. The methods by minimizing geometric distances have higher accuracies and robustness than other methods. However, the analytical difficulty in computing the geometric distances blocked the development of geometric fitting algorithms for general curves/surfaces for a long time. The project aims to develop a novel geometric distance between a point and a curve or a surface that allows easy computation and simultaneously for high accuracy. Based on the geometric distance, a robust cost function to fit the curve or surface will be established. The results are going to be applied to camera calibration. It is hopeful to have a flexible, easy, and high accurate calibration method that avoids establishing coordinate systems on which the popular calibration methods are dependent. Besides，the results would have many potential applications in computer graphics, robotics,3D printing and et al.
Keywords: Curve Fitting; Surface Fitting; Camera Calibration
Contact:
WU Yihong
E-mail: yhwu@nlpr.ia.ac.cn