基于标记点的多帧透视图像三维重建算法

3D Reconstruction from Multiple Perspective Views with Marked Points

提出了基于标记点的射影和欧氏重建方法,采用编码点和非编码点等标记点方式,利用编码点确定单CCD相机的外部姿态参数,用非编码点确定空间三维点的三维坐标。与已有方法比较,该方法有两个主要特点:①利用编码点进行加权迭代,由于编码点易于点对精确匹配,提高了算法的运算效率和重建精度;②将绝对二次曲面Ω的本质条件rank(Ω)=3施加到射影畸变矩阵中,提高了算法的鲁棒性。实验结果表明,利用该方法产生的三维重建点精确可靠,能够满足逆向工程等应用的要求。

This paper presented a method for 3D reconstruction from multiple perspective images. This method used the coded and the non-coded points. The coded points were utilized for posture recovery of a moving CCD camera and the non-coded points were utilized for 3D points reconstruction. Compared to previous methods, the novelties of the proposed method are two-fold. Firstly, the proposed method improves computing efficiency and reconstruction precision by weighted iterative implementing using the coded points being easy to match precisely. Secondly, the robustness is remarkably increased since the essential condition rank（Ω）= 3 of the absolute quadric is imposed to the projective distortion matrix. The experiment shows that the results are accurate and reliable to meet the requirements of reverse engineering applications.