基于基础矩阵的相对定向方法研究

针对倾角大的立体像对的相对定向不稳定等问题,提出了一种基于基础矩阵的相对定向方法,首先以SIFT匹配的同名点初步计算像对间的仿射变换,然后对影像分格网提取特征点并匹配出均匀分布的同名点,利用均匀分布的同名点通过奇异值分解方法解算基础矩阵参数,根据对应关系,将基础矩阵转化为立体像对的定向参数,最后通过相对定向严密解精确纠正定向参数。实验表明,本文的定向方法结果稳定,能够满足精度要求。

Study on Relative Orientation Method Based on Fundamental Matrix

In this paper, we propose a method of relative orientation based on fundamental matrix. Firstly, the affine transformation between image pair is calculated based on the corresponding points of SIFT matching. Then, the feature points are extracted from the image grid and matched with the corresponding points of uniform distribution. The fundamental matrix is transformed into the orientation parameters of the stereo image pair according to the corresponding relation. Finally, the directional parameters are corrected by the relative directional strict solution. The experiment shows that the directional method of this paper is stable to meet the accuracy requirements.