点云数据区域分割方法

为了便于逆向工程中模型的重建，针对散乱无组织的点云数据，提出了一种新的区域分割方法。首先对点云数据进行网格划分和稀化处理，对每个网格内的数据点用抛物面进行局部拟合，得到点云中每个点的法矢和曲率等微分信息，然后由“绝对标准”得到边界候选区域，再由“相对标准”得到最终边界点，连接这些边界点或对这些边界点进行曲线拟合，完成点云的区域分割。实践证明，通过该方法得到的边界点能满足工程应用的需要。

Segmentation Method for Point Cloud Data

To facilitate the shape reconstruction in reverse engineering,a new approach for segmenting scattered and unstructured point cloud data is presented.The raw point data set is first sorted into uniform 3-D grids,and the redundant points in every grid are filtered.Then these point data in each 3D-grid are fitted with a quadratic paraboloid to get differential properties for each point,such as: main curvature and normal vector value.The criteria of absolute value of peak curvature and bigger curvature over neighboring points are used to distinguish boundary points.An example is given to verify the presented method.