一种基于聚类的消失点自动测量方法

在人工环境中有很多平行的直线和相互正交的平面。本文提出了一种对相机拍摄的人工环境中的中远距离场景图像,自动获取其空间中三个方向相互正交的消失点的方法。利用中远距离场景中的平行直线在图像中倾角接近这一特点对图像中的边缘线段进行聚类初始化,并利用空间中平行直线在图像中相交于消失点这一事实进行聚类。三个消失点的三角形标准以及相机焦距标准将有助于消除虚假消失点。实验结果表明,该方法对中远距离场景的消失点检测可行、快速。

An vanishing points detection algorithm based on clustering

A man-made environment is characterized by many parallel lines and orthogonal edges.In this paper,a new algorithm is proposed for images of middle to long range scene taken from in the man-made environment in order to detect three vanishing points whose directions are mutually orthogonal in the real space.With edge lines extracted from the image,line clustering is initialized based on the fact that parallel lines in the middle to long range scene often have similar slopes in the image.Line clustering is then made based on the fact that parallel lines in the image will meet at the vanishing point respectively.The camera focus criterion and the triangle criterion of the three vanishing points will help to eliminate the falsely detected vanishing points.As shown in the experiments,the method is feasible and fast,and works automatically.