一种基于基本矩阵估计的立体视觉中滤除误匹配的方法

A Robust False Matches Filtering Method of Stereo Images Based on the Fundamental Matrix Estimation

根据基本矩阵建立两幅图像间的极线约束关系,能有效减少误匹配。噪声干扰和对应点中的误匹配使得基本矩阵的解精度降低。介绍了极线几何和基本矩阵理论,在最小中值平方法的基础上,提出一种基于匹配点对之间协因数的RANSAC(random sampling consensus)算法估计基本矩阵,有效解决了因误匹配导致的基本矩阵估计结果恶化问题。实验结果表明,所提出算法能有效滤除误匹配,具有良好的鲁棒性。

The false match can be eliminated with epipolar constraint relation set up between two images based on the fundamental matrix.Noise disturbance and correspondence outliers made the precision of the fundamental matrix very low.This paper introduced the theory of epipolar geometry and fundamental matrix and put forward a RANSAC algorithm to estimate the fundamental matrix based on the least median square method,effectively solving the deterioration problem for the fundamental matrix estimation induced by the false match.Experimental results prove that the algorithm proposed in the paper can effectively eliminate the false matches and is of good robust nature.