一种新的射影重建方法*

提出一种新的射影重建方法。它以最小化射影三维空间点的平均二维反投影误差为准则，以奇异值分解为工具，分步线性迭代实现求解过程，避免了传统射影重建方法中复杂的非线性优化过程和基础矩阵计算过程，且不需要任何初始估计，适用于匹配特征点存在“丢失点”的情况，不受相机特殊运动的限制。利用虚拟物体和真实物体图像序列进行了实验，证明该方法具有计算简单、准确性和鲁棒性高等方面的特点，具有较高的实用价值。

A New Method for Projective Reconstruction

A new method to recover projective reconstruction from multiple uncalibrated images is presented in this paper.Unlike the existing method,our algorithm proposes to estimate the projective reconstruction by minimizing the 2-D reprojection error,and it takes a linear iterative procedure in calculation with SVD(Singular Value Decomposition) as a main tool,consequently avoids the complex non-linear optimized process.In addition,no initial phase is needed to start the algorithm,and it can deal with the missing data in images and critical camera motion.Several experiments on simulated and real image data show that the proposed method is more accurate,robust and fast.