摘要
为解决目前点云精简算法适应性差的问题,提出一种基于特征显著性的自适应精简算法。通过对点云FPFH(fast point feature histograms)特征聚类生成特征单词;在考虑单词间差异的基础上,融合单词内部的特征分散程度,形成显著性词典,由词典软编码单点特征,得到点云特征显著性;在均匀网格基础上,若网格内的特征显著性越强,则配置越高的采样率,由此实现点云的自适应精简。实验结果表明,所提算法能够区分出点云中的特征明显区域,在精简不同尺寸、形状点云时具有适应性。
To solve the problem of poor adaptability of current feature-based point cloud simplification,an adaptive simplification of point cloud based on feature saliency was proposed.Feature words were generated by clustering point cloud FPFH(fast point feature histograms)features.A saliency dictionary was formulated considering not only the dissimilarity between different words but also the feature divergence with the word.According to the saliency dictionary,each point feature was softly encoded into a saliency value,namely the point feature saliency.After uniformly meshing the point cloud,the stronger the feature saliency in the grid was,the higher the sampling rate was configured.Experimental results show that the proposed algorithm can effectively distinguish the obvious feature areas in the point cloud,and it is adaptive to simplify point cloud with different sizes and shapes.
作者
张亦芳
李立
刘光帅
ZHANG Yi-fang;LI Li;LIU Guang-shuai(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)
出处
《计算机工程与设计》
北大核心
2021年第8期2211-2217,共7页
Computer Engineering and Design
基金
国家自然科学基金项目(51275431)
中国电子科技集团公司第十研究所技术创新基金项目(十所计20181218)。
