采用密度空间聚类的散乱点云特征提取方法

发布时间：2018-01-19 01:23

本文关键词： 点云特征提取基于密度空间聚类全局约束性反k近邻　出处：《西安电子科技大学学报》2017年02期 　论文类型：期刊论文

【摘要】：针对现有点云特征提取算法中,采用全局特征度量阈值及仅使用点的局部信息进行特征提取而造成的特征尖锐程度敏感、对潜在曲面差异较大模型效果差等问题,提出一种基于密度空间聚类的散乱点云特征提取方法.首先,对点的反k近邻进行定义,并提出一种新的特征检测算子;然后,将反k近邻的尺度作为点密度,引入特征的全局约束性信息;最后,对基于密度空间聚类方法中的相关概念进行重定义并建立新的特征识别准则,提取特征点.实验结果表明,该算法简单、有效、鲁棒,同时考虑了特征的局部性信息及全局约束性信息,针对潜在曲面形状差异较大的模型表现出了较强的优越性.
[Abstract]:In the existing point cloud feature extraction algorithms, the global feature measurement threshold and the feature extraction using only the local information of the point are sensitive to the feature sharpness. In order to solve the problem of large difference of potential surfaces, a method of feature extraction of scattered point clouds based on density space clustering is proposed. Firstly, the inverse k-nearest neighbor of points is defined. A new feature detection operator is proposed. Then, the scale of the inverse k-nearest neighbor is taken as the point density, and the global constraint information of the feature is introduced. Finally, the related concepts in the density-based clustering method are redefined and a new feature recognition criterion is established. The experimental results show that the algorithm is simple, effective and robust. At the same time, the local information and the global constraint information of the feature are taken into account, and the model with large difference in the shape of the potential surface is shown to be superior.
【作者单位】：西北大学信息科学与技术学院;
【基金】：国家自然科学基金资助项目(61373117,61572400);国家自然科学基金青年基金资助项目(61305032) 陕西省教育厅科研专项资助项目(2013JK1180)
【分类号】：TP311.13;TP391.7
【正文快照】： 三维扫描技术的出现使得快速方便地获取能表达现实物体表面形状的点云模型成为可能.点云模型处理已成为几何造型领域内的研究热点,并广泛应用于医学、艺术、模式识别以及文物数字化保护等领域[1-3].特征点是最基本的曲面几何形状的特征基元,对于几何模型的外观及其准确表达具

【参考文献】