摘要： 以基础矩阵估计规范八点算法为基础,针对线性算法对参考图像匹配点集中存在的异常匹配极为敏感的问题,将基础矩阵的线性鲁棒估计转换为对匹配点集的最优化采样问题,以最小对极距离平方和中数准则获得基础矩阵的稳定可信解。实验表明,算法对基础矩阵的精度和稳定性均有较大提高。

关键词: 基础矩阵, 异常匹配, 匹配点集采样, 鲁棒估计

Abstract: The fundamental matrix estimation arithmetic is sensitive to the outlier match. This paper proposes a robust arithmetic based onnormalized 8-point arithmetic which focus on the match set sampling. The least median of squares theory is applied to the match set optimizationand the fundamental matrix is estimated on the minimized solvable subset. Experiments show that the new arithmetic is very robust to noise andmismatches, and the fundamental matrix with high accuracy can be found.

中图分类号: 

• TP391.41