3D Multi-Angle Point Cloud Stitching Using Iterative Closest-point Stitching and K-Nearest-Neighbors
3D Multi-Angle Point Cloud Stitching Using Iterative Closest-point Stitching and K-Nearest-Neighbors
复制标题
DOI：
10.1109/iccsi55536.2022.9970689
复制
发表时间：
2022-11
期刊：
影响因子：
--
通讯作者：
Pankti Patel;Ryan Hare;Ying Tang;Nidhi Patel
中科院分区：
文献类型：
--
作者：
Pankti Patel;Ryan Hare;Ying Tang;Nidhi Patel
最近对虚拟环境和3D对象扫描的关注突出了对将并发点云缝合到实体三维（3D）模型中的准确且高效的方法的需求。为了满足这一需求，我们引入了一种新的迭代方法，用于3D多角度点云拼接，使用迭代最近点（ICP）算法与k-最近邻（kNN）增强。通过这种组合算法，我们的方法专注于最大限度地减少相邻点云之间的误差，使我们能够轻松地计算必要的变换，将联合收割机点云组合成一个模型。因此，当给定在同一对象的多个角度捕获的并发点云时，我们的方法提供了单个准确的3D模型。我们评估了所提出的框架的能力，缝合多个点云到一个实体模型，通过缝合分割模型和比较均方根误差的标准迭代最近点缝合算法。实验结果表明，与标准方法相比，我们的方法在效率和准确性方面具有优势。
The recent focus on virtual environments and 3D object scanning has highlighted the need for accurate and efficient methods to stitch concurrent point clouds into solid three-dimensional (3D) models. To address this need, we introduce a novel iterative approach for 3D multi-angle point cloud stitching using an iterative closest point (ICP) algorithm augmented with k-nearest neighbors (kNN). With this combined algorithm, our method focuses on minimizing the error between neighboring point clouds, allowing us to easily compute the necessary transformation to combine point clouds into one model. Thus, when given concurrent point clouds captured at multiple angles of the same object, our approach provides a single accurate 3D model. We evaluated the ability of the proposed framework to stitch multiple point clouds into a solid model by stitching a segmented model and comparing the root mean squared error to a standard iterative closest-point stitching algorithm. The experiments results shows that our method provides benefits in terms of efficiency and accuracy compared to a standard approach.