基于RGB-D相机的多视角机械零件三维重建

信寄遥; 陈成军?覮; 李东年

基于RGB-D相机的多视角机械零件三维重建

引用本文：信寄遥，陈成军?覮，李东年.基于RGB-D相机的多视角机械零件三维重建[J].计算技术与自动化,2020,(3):147-152

摘要点击次数: 208

全文下载次数: 0

作者	单位
信寄遥，陈成军?覮，李东年	（青岛理工大学机械与汽车工程学院，山东青岛 266520）

中文摘要:为了低成本且高效的实现对机械零件的三维重建和参数测量，研究了利用RGB-D相机从6个角度拍摄机械零件，获得零件不同角度的深度图像与彩色图像，通过坐标转换将深度图像转换成点云数据。首先利用滤波算法去除点云噪声，分割出机械零部件的点云数据，并利用PCA主成分分析法计算点云数据的法向量；使用最近点迭代算法（ICP）实现相邻三视角点云数据的配准，得到正背面的点云，将正背面点云进行旋转融合得到最终的目标点云数据；最后使用泊松重建算法得到完整闭合的零件三维模型。实验结果表明本文的三维重建方法具有较好的鲁棒性和准确性，重建得到的三维模型细节清晰，点云误差较小。

中文关键词:三维重建点云配准模型重构最近点迭代算法 Kinect

3D Reconstruction of Mechanical Parts Using Multi-view RGB-D Images

Abstract:In order to realize the 3D reconstruction and measurement of surface parameters of mechanical parts with low cost and high efficiency，the mechanical parts are taken from 6 angles by RGB-D camera，and the depth and color images of different angles of mechanical part are obtained，and the depth image is converted to point cloud data by coordinate transformation. firstly，the filtering algorithm is used to remove the point cloud noise，and segment the point cloud data of the mechanical components. And Principal Component Analysis （PCA） is ued to calculate the normal vector of the point cloud data. Secondly，by the Iterative Closest Points（ICP）method，the registration of the adjacent three-point point cloud data is realized. So the point cloud on the back side，is rotated and fused to obtain the final target point cloud data. Finally，the Poisson reconstruction algorithm is used to obtain the 3D model. The experimental results show that the proposed 3D reconstruction method has better robustness and accuracy，and the reconstructed 3D model has clear details and less point cloud error.

keywords:3D reconstruction point cloud registration model reconstruction iterative closest points kinect

查看全文 查看/发表评论 下载pdf阅读器