.png){kind=icon}
Return
Efficient multi-scale 3D point cloud registration
.png){kind=icon}
PRE
AI
DOI：10.1007/s00607-024-01401-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study introduces DMR-PointHop, standing for deep learning-based multi-resolution PointHop, a novel multi-scale fusion-based method for 3D point cloud registration. The method leverages multi-level downsampling to create datasets of varying scales and establishes local reference coordinate systems for feature extraction at each point. The extraction, fusion, and hierarchical learning of features across different scales enable DMR-PointHop to capture information from local to global levels. Subsequently, it identifies a salient subset of points with reliable correspondence for accurate point cloud matching prediction. The method utilizes singular value decomposition to compute rigid transformations, ensuring precise registration with robustness to rotation and translation. The study demonstrates that DMR-PointHop contributes to advancements in artificial intelligence by providing a more efficient and environmentally friendly point cloud registration solution compared to traditional methods and some deep learning approaches.
Keywords:
Point cloud registration
Multi-scale fusion
Rotation invariance
3D descriptor
Journal
C
IF:
2.8
Papers:
2.3K
Citations:
3.5K
Organization
Cited Papers
ShareSave
ShareSaveSubversion of Autophagy in Adherent Invasive Escherichia coli-Infected Neutrophils Induces Inflammation and Cell Death.png)
PLoS ONE
OAAI
PLoS ONEIF0
OAAIShareSaveShareSaveShareSaveShareSave
View more