基于参数自适应各向异性高斯核的散乱点云保特征去噪

发布时间：2018-04-22 14:23

本文选题：信息相似度 + 有效邻域　；参考：《计算机集成制造系统》2017年12期

【摘要】：为解决传统点云去噪算法造成的过光顺及局部失真问题,提出一种基于各向异性高斯核的散乱点云保特征去噪算法。根据邻域点在其切平面上的投影和采样点法向构建信息相似度函数,并通过信息相似度函数定义有效邻域;应用主元分析理论研究曲面采样点、棱线采样点和角点的特征值及特征向量的分布特性;在此基础上构建以协方差矩阵的伪逆矩阵为带宽矩阵的参数自适应的各向异性高斯核函数,并将其与双边滤波算法结合用于散乱点云去噪。实验结果表明,该算法能够根据点云的局部分布特性自适应地调整滤波主方向和各主方向的衰减速度,在实现散乱点云去噪的同时可有效保持点云模型的原始尖锐特征。
[Abstract]:In order to solve the problem of over-fairing and local distortion caused by the traditional point cloud denoising algorithm, a feature preserving denoising algorithm for scattered point cloud based on anisotropic Gao Si kernel is proposed. According to the projection of neighborhood points on its tangent plane and the normal direction of sampling points, the information similarity function is constructed, and the effective neighborhood is defined by the information similarity function. On the basis of the eigenvalues and the distribution of eigenvector of prism sampling point and corner point, an anisotropic Gao Si kernel function with the parameter adaptive of the pseudo inverse matrix of covariance matrix as bandwidth matrix is constructed. It is combined with the bilateral filtering algorithm for scattered point cloud denoising. The experimental results show that the algorithm can adaptively adjust the attenuation speed of the main direction and each direction of the filter according to the local distribution characteristics of the point cloud, and it can effectively preserve the original sharp feature of the point cloud model while realizing the de-noising of the scattered point cloud.
【作者单位】：燕山大学电气工程学院;哈尔滨市科佳通用机电股份有限公司;
【基金】：国家自然科学基金资助项目(51305390,61501394) 河北省自然科学基金资助项目(F2016203312)~~
【分类号】：TP391.41

【相似文献】