Li Xurui, Liu Guangshuai, Sun Si
J Opt Soc Am A Opt Image Sci Vis. 2021 Jan 1;38(1):60-70. doi: 10.1364/JOSAA.410458.
Efficient and quick extraction of unknown objects in cluttered 3D scenes plays a significant role in robotics tasks such as object search, grasping, and manipulation. This paper describes a geometric-based unsupervised approach for the segmentation of cluttered scenes into objects. The proposed method first over-segments the raw point clouds into supervoxels to provide a more natural representation of 3D point clouds and reduce the computational cost with a minimal loss of geometric information. Then the fully connected local area linkage graph is used to distinguish between planar and nonplanar adjacent patches. Then the initial segmentation is completed utilizing the geometric features and local surface convexities. After the initial segmentation, many subgraphs are generated, each of which represents an individual object or part of it. Finally, we use the plane extracted from the scene to refine the initial segmentation result under the framework of global energy optimization. Experiments on the Object Cluttered Indoor Dataset dataset indicate that the proposed method can outperform the representative segmentation algorithms in terms of weighted overlap and accuracy, while our method has good robustness and real-time performance.
在诸如物体搜索、抓取和操作等机器人任务中，高效快速地从杂乱的三维场景中提取未知物体起着重要作用。本文描述了一种基于几何的无监督方法，用于将杂乱场景分割成物体。所提出的方法首先将原始点云过度分割成超体素，以提供三维点云更自然的表示，并在几何信息损失最小的情况下降低计算成本。然后使用全连接局部区域链接图来区分平面和非平面相邻面片。接着利用几何特征和局部表面凸度完成初始分割。初始分割后，会生成许多子图，每个子图代表一个单独的物体或其一部分。最后，我们使用从场景中提取的平面在全局能量优化框架下细化初始分割结果。在物体杂乱室内数据集上的实验表明，所提出的方法在加权重叠和准确性方面优于代表性的分割算法，同时我们的方法具有良好的鲁棒性和实时性能。
