基于改进FPFH特征提取的点云模板匹配方法
Point Cloud Template Matching Method Based on Improved FPFH Feature Extraction
赵云涛 1胡佳明 2李维刚 1谢万琪2
作者信息
- 1. 武汉科技大学冶金自动化与检测技术教育部工程研究中心,湖北武汉 430081;武汉科技大学信息科学与工程学院,湖北武汉 430081
- 2. 武汉科技大学信息科学与工程学院,湖北武汉 430081
- 折叠
摘要
利用传统的配准算法在对源点云和场景目标点云进行点云配准时,由于环境光照和设备精度的影响,使得捕获的场景点云往往存在大量噪声,并且场景数据捕获不全,导致配准精度低、迭代速度慢,容易产生局部最优解等问题.针对该情况,提出一种基于改进快速点特征直方图(FPFH)特征提取的点云模板匹配方法.该方法利用移动最小二乘法(MLS)平滑波动数据、修补点云孔洞的特点,从而提高FPFH描述子的特征权重,优化源点云和目标点云的对应关系.最后利用对应关系进行采样一致性初始配准(SAC-IA)和迭代最近点(ICP)配准得到最终的变换矩阵.通过对比实验验证了该方法相比传统的配准算法在场景点云中对目标点云进行模板匹配时迭代次数降低了 35％,配准精度提高了 82％,具有配准精度高、鲁棒性强、可靠性高的特点.
Abstract
Challenges in point cloud registration arise from noise in captured scene data,often due to environmental lighting and equipment limitations,leading to incomplete scene capture and issues such as low accuracy,slow iteration,and susceptibility to local optima.To address these issues,we propose a point cloud template matching method that leverages an enhanced Fast Point Feature Histogram(FPFH)feature extraction technique.This method uses the characteristics of moving least squares(MLS),which can smooth the fluctuation data and repair the holes in the point cloud,so as to improve the feature weight of the FPFH descriptor and optimize the corresponding relationship between the source point cloud and the target point cloud.Finally,the corresponding relationship is used for the initial registration of sampling consistency(SAC-IA)and the iterative closest point(ICP)registration to obtain the final transformation matrix.This approach demonstrates a 35％reduction in itera-tion count and an 82％improvement in registration accuracy compared to traditional algorithms when matching the target point cloud within the scene data.The method exhibits high accuracy,robustness,and reliability in point cloud registration tasks.
关键词
机器视觉/快速点特征直方图/特征提取/移动最小二乘法/点云配准Key words
machine vision/fast point feature histogram/feature extraction/moving least square method/point cloud registration引用本文复制引用
出版年
2024
NETL
NSTL
万方数据