Detection of co-planar circle pair of same radius from a single image
Detection of co-planar circle pair of same radius from a single image
复制标题
DOI：
10.1109/robio.2016.7866474
复制
发表时间：
2016-12
期刊：
影响因子：
--
通讯作者：
Mizokami Naoki;Haiyuan Wu;Qian Chen;Kazumasa Suzuki;R. Sakamoto
中科院分区：
文献类型：
--
作者：
Mizokami Naoki;Haiyuan Wu;Qian Chen;Kazumasa Suzuki;R. Sakamoto
本文提出了一种从复杂背景图像中检测出两个半径相同的共面圆形目标的方法。我们提出了一种独特的基于RANSAC的方法来检测椭圆的候选人，该方法将以假设的椭圆中心为中心的圆形区域划分为N个相等的扇区，然后从每个扇区中随机选择边缘点进行拟合。我们提出了三个几何约束的一对共面圆具有相同的半径，考虑从每个椭圆和两个圆的中心的三维坐标估计的支持平面的法向量。从检测到的候选椭圆，我们选择具有相同半径的共面圆对，通过检查它们是否符合所提出的几何约束。该方法的输出可以用作许多需要平行或共面圆的相机校准和/或3D结构恢复算法的输入。我们的方法也可以应用到移动的机器人，以确定他们的位置，通过监测环境中的两个共面的圆形物体。
This paper describes a method that detects two co-planar circular shaped objects having the same radius from an image with clutter background. We propose an unique RANSAC based method for detecting the candidates of ellipses that divides the circular region centered at the assumed ellipse center into N equal sectors, then fits the edge points selected randomly from each sector. We propose three geometrical constraints of a pair of co-planar circles having the same radius by considering the normal vectors of the supporting plane estimated from each ellipse and the 3D coordinates of the centers of the two circles. From the detected candidates of ellipses, we select the co-planar circle pairs having the same radius by checking how well they fit the proposed geometrical constraints. The output of this method can be used as inputs of many camera calibration and/or 3D structure recovery algorithms that require parallel or co-planar circles. Our method can also applied to the mobile robots to determine their position by monitoring the two co-planar circular objects in the environment.