公开(公告)号：KR1020140087235A

公开(公告)日：2014-07-09

申请号：KR1020120156585

申请日：2012-12-28

Applicant: 성균관대학교산학협력단

Inventor： 이석한 ,  김재웅 ,  후훙뉘엔

IPC: G06F17/00

CPC classification number: G06F17/16 ,  G06T3/0037

Abstract: A planar surface patch plays a key functional role in recognizing an entity and a condition. As an important technical issue associated with the extraction side of such a planar surface patch, there still is a tradeoff between accuracy in modeling and efficiency in operation. The present invention presents a method of acquiring high accuracy in modeling from less operation and deals with a method of extracting a planar surface patch from 3D point clouds having many errors. An important contribution provided by a proposed method is as follows. 1) The strong estimation of a surface normal vector may minimize an influence on data noise. 2) As a method of identifying a planar surface patch representing a similar direction, there is a determination method of the variable size of a sliding window used in estimating the accurate location of a peak having a high density and in estimating a general vector density. The present invention shows an enhanced result in comparison typical methods through experiment results on various interior and exterior images.

Abstract translation: 平面表面贴片在识别实体和条件方面起着关键的功能作用。 作为与这种平面表面贴片的提取侧相关联的重要技术问题，在建模精度和操作效率之间仍然存在折衷。 本发明提出了一种从较少的操作中获得高精度建模的方法，并且涉及从具有许多误差的3D点云提取平面表面贴片的方法。 提出的方法提供的重要贡献如下。 1）表面法向量的强估计可以最小化对数据噪声的影响。 2）作为用于识别表示相似方向的平面表面贴片的方法，存在用于估计具有高密度的峰值的精确位置并用于估计通用矢量密度的滑动窗口的可变大小的确定方法。 本发明通过各种内部和外部图像的实验结果比较典型的方法来显示增强的结果。