基于经验模态分解的点云数据平滑与增强

发布时间：2019-05-12 01:28

【摘要】：在计算机辅助设计与逆向工程应用中,针对缺乏拓扑连接关系的点云数据,提出了基于经验模态分解(EMD)的点云数据平滑与增强算法。首先,以点云模型的拉普拉斯矩阵坐标与法向的内积作为EMD输入信号,提取点云模型输入信号的极值点作为插值节点计算信号的上下包络;然后,为实现特征保持的EMD信号分解,通过检测点云数据上特征点,并在计算信号上下包络的过程中作为约束,克服传统EMD算法无法保持特征的局限;最后,迭代地从输入信号中减去上下包络的均值得到内蕴模态函数(IMF)和余量,并通过设计滤波器实现了点云数据平滑和增强。实验结果表明,本文算法有效地将EMD推广到三维散乱点云数据中,扩大EMD在三维几何中的应用范围,并在点云数据平滑和增强方面取得了很好的效果。
[Abstract]:In the application of computer aided design and reverse engineering, a point cloud data smoothing and enhancement algorithm based on empirical mode decomposition (EMD) is proposed for point cloud data which lack topological connection. Firstly, the Laplacian matrix coordinates and the normal inner product of the point cloud model are used as the EMD input signal, and the extreme point of the input signal of the point cloud model is extracted as the interpolation node to calculate the upper and lower envelopes of the signal. Then, in order to decompose the EMD signal with feature preservation, the feature points on the point cloud data are detected and used as constraints in the process of calculating the upper and lower envelope of the signal to overcome the limitation that the traditional EMD algorithm can not maintain the feature. Finally, the intrinsic modal function (IMF) and margin are obtained iteratively by subtracted the mean value of the upper and lower envelopes from the input signal, and the point cloud data are smoothed and enhanced by designing a filter. The experimental results show that the algorithm effectively extends EMD to 3D scattered point cloud data, expands the application range of EMD in 3D geometry, and achieves good results in point cloud data smoothing and enhancement.
【作者单位】：北京航空航天大学计算机学院;天津工业大学理学院;
【基金】：国家自然科学基金(61532002,61672149,61602341,11626169) 天津市自然科学基金(17JCQNJC00600) 虚拟现实技术与系统国家重点实验室(北京航空天大学)开放基金(BUAA-VR-17KF-04)~~
【分类号】：TN911.7

【相似文献】