Fei Xia,∗ Jiale Chen,∗ Yu Pan,∗ Wei Zhang,∗∗ Wantao Zhuo,∗ and Jianliang Mao∗
K-Means, particle swarm optimisation (PSO), point cloud denoising, gravity, gravitational feature function
Point cloud noise adversely affects the accuracy of 6D grasping pose estimation, particularly when generated by depth cameras. As a result, the denoising process is essential for the successful operation of robotic manipulators during grasping tasks. To address this issue, a 3D point cloud denoising algorithm using optimised K-Means clustering and gravitational feature functions is proposed, with the optimal K value determined through a combination of particle swarm optimisation PSO and the elbow method. The gravity centre and gravitational value for each point cloud type are computed in 3D space. Based on these values, gravitational feature functions facilitate noise removal according to gravitational laws. The precision is improved by 3%–27% and the recall is enhanced by 15%–47% compared to traditional K-Means, DBSCAN, Optics, and Spectral algorithms, as evaluated on the Bunny and Happy Buddha datasets with 20% added random noise.
Important Links:
Go Back
