一种基于L_∞范式的求解已知旋转相机重建算法

基于多幅图像序列的三维重建过程中,相机模型的坐标统一是非常重要的基础。在已知两两相机之间的相对关系的情况下,将相机模型统一至同一世界坐标系,并恢复特征点的三维坐标。本文给出了一种基于L∞范式的求解已知旋转相机重建方法。给出了一种基于L∞范式的几何结构和运动问题的新框架,并用实验证明了算法能够达到很好的性能。

Designing Bézier surfaces minimizing the L^2-norm of the Gaussian curvature

In freeform surface modelling, developable surfaces have much application value. But, in 3D space, there is not always a regular developable surface which interpolates the given boundary of an arbitrary piecewise smooth closed curve. In this paper, tensor product Bézier surfaces interpolating the closed curves are determined and the resulting surface is a minimum of the functional defined by the L2-integral norm of the Gaussian curvature. The Gaussian curvature of the surfaces is minimized by the method of solving nonlinear optimization problems. An improved approach trust-region form method is proposed. A simple application example is also given.