基于ABM与FBM结合的精确快速相机自标定方法

Quick and accurate camera self-calibration method based on combination of FBM and ABM

针对BA在场景重构时精度达不到摄影测量规范要求这一问题,提出一种SFM流程,旨在最大化增量重建的精度。采用分层匹配方法,在图像金字塔顶层进行基于特征点的匹配(FBM),采用带有改进的NCC和LSM的基于区域匹配(ABM)方法逐层对匹配点位置进行调整,得到高精确度的匹配位置;将修正后的匹配点用于基础矩阵F和单应矩阵H的估计,计算tracks信息,生成图像连通图G,用于决定图像添加到BA中的顺序,从而得到高精度的相机位姿和3D点。实验结果表明,该流程能够提高重建结果的精度并实现可扩展性。

Aiming at the problem that the accuracy fails to satisfy the requirements of photogrammetry specifications when recon- structing the scene by BA （bundler adjustment）, a new process of SFM （structure from motion） was proposed aiming to maxi- mize the accuracy of the incremental reconstruction. A hierarchical matching method was used and the feature based matching （FBM） was run at the top level, then the position of matching points were adjusted layer-by-layer using area based matching （ABM） with improved normalized cross correlation （NCC） and least squares image matching （LSM） to obtain the high accuracy matching position. Corrected points were used to estimate the fundamental matrix and homography matrix, and tracks were computed and the image connectivity graph G was constructed. Tracks and G were used to determine the sequence of images added into BA to obtain high precision camera poses and 3D points. Results of experiments show that the proposed method can improve the precision of reconstruction and achieve sealability.