点云数据模型特征点提取算法研究与实现

Research and implementation of algorithm for point cloud data model and feature point extraction

以LS方法为基础,在对已有的特征点提取方法研究的基础上,设计了一种改进的点云模型特征点提取算法。该方法以扫描到的每个数据点为中心建立一个δ-邻域,再以LS方法为基础,拟合此邻域的一个微切平面,然后将邻域内的点透视投影到此微切平面,计算这些点在微切面上的投影点的协方差值,从而构造一个协方差矩阵,并通过计算矩阵的相关值,设定阈值,把满足一定阈值范围内的点作为研究模型的边界特征点。

The paper is based on the LS （Least square ） method, considering the advantages and dis- advantages of other feature points. And it designed an algorithm of an improved point cloud model of feature points. Firstly, each data points to scan to the establishment of a delta-Neighborhood Center. Then, on the basis of LS method, a tangent plane fits this neighborhood. Secondly, it put these neigh- borhood points perspective projects on the tangent plane and calculate the covariance value of these projection points , so as to construct a collaborators covariance matrix, and calculate the relative value of the matrix. Finally, it sets a threshold ,meeting a certain threshold range points as the boundary fea- ture points of model.