文中提出了一种新的基于弱平行关系的灭点估计算法.算法的框架如下:在单幅图像中标记所有弱平行线段并计算其对应的弱平行关系度量值;选择那些具有较大弱平行关系度量值的弱平行线段,在图像上找到它们的伪灭点并用高斯分布模型和交叉迭代方法对选取的伪灭点进一步精确其对应的弱平行关系度量值;选择精化后的拥有较大弱平行关系度量值的伪灭点,用最小二次拟合方法来得到无穷远直线在图像平面上的投影直线.该方法避免了基于灭点的三维重建算法中要求先验得到平行关系的条件,而是以场景中更常见的无需先验知道的弱平行关系为三维重建基础,使算法具有了更广适用性和更好稳定性.
The authors propose a novel algorithm for computing vanishing points from based on weak-parallel line segments and a concept of weak-parallel. The framework of t as follows, firstly, the line segments are obtained which are weak-parallel in the image, urements of weak-parallel between them is calculated; secondly, the pairwise line segm single image algorithm is en ents the meas- with large weak-parallel measurements are chosen to find the pseudo-vanishing points, and the points are confirmed by using Gaussian probability model and Cross-iteration method; finally, the pseudo-vanishing points with large weak-parallel measurements are selected to locate the projection of infinity line in the image plane. Compared to the 3-dimensional reconstruction algorithms partly based on strict parallel lines, the proposed algorithm is able to be used more widely and stably.