基于曲线拟合和SIFT算子的点云关键点提取  

Point Cloud Key Point Extraction Based on Curve Fitting and SIFT Operator

作　　者：曹烨 郭丽红[1] 单锦涛 纪宇菲 胡婷婷[1] Cao Ye

机构地区：[1]南京工程学院信息与通信工程学院,江苏南京211167

出　　处：《工业控制计算机》2024年第11期45-46,49,共3页Industrial Control Computer

基　　金：江苏省大学生科技创新项目(202311276062Y);南京工程学院校级基金项目(YKJ202214)。

摘　　要：针对传统点云关键点提取算法中存在的关键点缺失问题,以槽型零件的切面点云为例,提出一种基于曲线拟合和SIFT算子的点云关键点提取算法。由于平面点云数据间特征关联性较差且极易受到离群点的干扰,因此考虑从点云的图像化处理着手,首先通过滤波算法实现点云的预处理;然后,采用决策树拟合的方式实现点云数据到图片的转换;随后,通过SIFT算子提取到点云关键点;最后,依据数据到图片的转换关系完成点云关键点的校准。通过实验可知,点云数据的图像化处理方法能弱化点云数据中的不明显离群点和褶皱处的干扰,能高效准确地实现点云关键点的提取。In order to solve the problem of missing key points in traditional point cloud key point’s extraction algorithm,a new point cloud key points extraction algorithm based on curve fitting and SIFT operator is proposed in this paper.Because of the poor correlation of the features among the point cloud data and the easy interference of outliers,the point cloud image processing is considered.Firstly,the point cloud is pre-processed by filtering algorithm,then the method of decision tree fitting is adopted to realize the conversion of point cloud data to picture.Next,the key points of point cloud are extracted by SIFT operator.Finally,the key points of the point cloud are calibrated according to the conversion relationship between data and picture.The experiment shows that the operation of point cloud image can weaken the interference of non-obvious outliers and folds in point cloud data and extract the key points efficiently and accurately.

关 键 词：关键点提取 决策树拟合 SIFT算子 校准算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]