基于散乱点云的多尺度特征点提取算法  

Multi-scale feature point extraction algorithm based on scattered point cloud

作　　者：蔡晋茹 况立群[1] 韩燮[1] 

机构地区：[1]中北大学计算机与控制工程学院,山西太原030051

出　　处：《计算机工程与设计》2016年第12期3255-3259,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61379080);山西省自然科学基金项目(2014011018-3);国家科技支撑计划基金项目(2013BAH45F02)

摘　　要：针对物体的尺度不变性特征以及传统算法对噪声敏感等问题,本文提出一种基于散乱点云的多尺度特征提取算法。通过改变局部邻域的大小,构造尺度空间,进行多尺度分析,在不同的尺度下,通过对局部邻域的协方差分析,计算曲面变化值,找到具有尺度不变性的特征点;引入基于形状索引值的点签名方法,增强对噪声的鲁棒性。Aiming at the problem of noise sensitivity of traditional algorithm and for the scale invariant feature,a multi-scale feature extraction algorithm based on scattered point cloud was proposed.Multi-scale analysis of scale space was constructed by changing the local neighborhood size.In the different scales,the variation of the surface was calculated by the covariance analysis of the local neighborhood,and the feature points with scale invariance were found.At the same time,the point signature method was introduced to enhance the robustness to the noise.

关 键 词：散乱点云 尺度空间 协方差分析 特征点提取 点签名 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]