一种新的点云数据特征骨架提取方法

A new method for extracting feature skeleton from point cloud

为解决点云数据的线骨架提取问题,为点云数据的后续几何处理的奠定基础,提出了一种新的点云数据骨架提取方法.通过对点云数据的空间层次剖分后建立其简化模型,可有效地避免噪声点对骨架的干扰;根据离散Morse理论,从简化模型中提取主要的特征点,用测地线连接这些主要特征点可得到模型的初步骨架.采用可见反力场方法将初步骨架内推至模型内部,对内推后的骨架光顺及聚类后形成最终骨架.该方法能够直接处理带噪声数据的大规模点云数据,所形成的骨架连续.

A simplified geometric model was constructed for point clouds, which is not sensitive to noises. Then feature points were found from the simplified model by Morse theory. Connecting those features by geodesic lines, an initial skeleton could be obtained, which were pushed inside the point cloud by visible repulsive forces. The skeletons were then smoothed and clustered with angle threshold to form a final skeleton. The method can process large point cloud model with noises, and the extracted skeletons are continuous.