一种半自动的三维欧氏重建方法

从二维图像得到场景的三维模型是计算机视觉和虚拟现实的重要研究内容。本文通过用户的简单交互,利用平面型场景的同形关系自动进行两幅大差异图像之间的角点匹配,将这些匹配结果作为初始点,再利用RANSAC鲁棒算法估计基本矩阵。以此结果进行仿射重建,然后在简化相机模型的基础上通过给出的约束条件直接实现欧氏重建,真实数据的实验结果表明了该算法的有效性。

A Semiautomatic Method of 3D Euclidean Reconstruction

3D reconstruction from 2D images is one of the most important research issues in computer vision and virtual reality.By means of simple interaction,our method can get the homography matrix and automatic matching between two wide-baseline images.After getting the initial matching points,we have estimated the fundamental matrix using robust RANSAC method.Firstly,we have gotten a projective reconstruction and then a Euclidean reconstruction based on simplified camera model using given constraints.Results of real image data demonstrate this method.