影像边缘特征与LS-SVM的点云边缘残缺区域修补方法

针对点云修补过程中点云边缘的残缺区域边界信息的不确定性问题,本文提出了一种基于影像边缘特征与LS-SVM的点云边缘残缺区域修补方法:首先将影像与点云进行配准,并利用亚像素边缘检测算法提取目标边缘特征;然后构造一特征平面,同时将训练样本集与目标边缘特征投影至该平面,以确定重采样范围与点位;通过利用最小二乘支持向量机回归方法,获得残缺区域的曲面方程并进行重采样,最终完成修补。实验证明,该方法得到的修补点云与原始数据融合平滑,修补效果符合实际目标的特征。

Fragmentary area repairing on the edge of point clouds based on image edge extraction and LS-SVM

In the process of hole-repairing in point clouds,it’s difficult to repair by the indeterminate boundary of fragmentary area in the edge of point clouds.In view of this condition,the article advanced a method of fragmentary area repairing on the edge of point clouds based on image edge extraction and LS-SVM.After the registration of point clouds and corresponding image,the sub-pixel edge could be extracted from the image.Then it projected the training points and sub-pixel edge to the characteristic plane to confirm the bound and position for re-sampling.At last it got the equation of fragmentary area to accomplish the repairing by Least-Squares Support Vector Machines.The experimental results demonstrated that the method could guarantee accurate fine registration.