多视点标定图像的交替迭代度量重建方法

提出了一种针对相机内部参数已标定的多视点图像交替迭代度量重建方法.该方法以目标空间中重建的3D点到匹配特征点的反投影线间的距离作为重建误差的测度,使误差函数形式上更简洁且具有明确的几何意义.重建过程通过对相机外部参数中的旋转矩阵、平移量和3D重建点交替迭代完成,其中平移量和3D点一起估计,避免了它们与旋转矩阵中的元素间的尺度差异引起的数值偏差.利用合成数据和真实图像的计算分析均表明本文方法有较高的重建精度,具有实用性.

Alternating iterative metric reconstruction from calibrated multi-view images

An alternating iterative method for 3D metric reconstruction from calibrated multi-view images is proposed in this paper.It is based on the error distance which is defined as the length from the reconstructed 3D point to the ray passing through the matched feature point and the camera center.The above error distance is simple and meaningful.The 3D reconstruction is completed by alternating iteration of translation vectors,3D points and rotation matrices.The translation vectors and 3D points are estimated simultaneously to avoid the numerical bias raised from the different scale of the optimization variables.The experiments on synthetic data and real images show that the proposed method has high accuracy and can be used in practice.