Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

GLLA: A Unified Force-Directed Graph Layout Framework Supporting Local Adjustments.

IEEE transactions on visualization and computer graphics·2026
Same author

CIRBP exerts neuroprotective effects in ischemic injury by inhibiting IFN-I signaling.

Life sciences·2026
Same author

Tumor cell extracellular vesicles derived lncRNA XIST regulates the blood-brain barrier and NSCLC brain metastasis via miR-19b-3p/EPN2.

Brain research·2026
Same author

Microbial community reshape and TTMP enhancement in daqu via targeted bioaugmentation of Bacillus velezensis and Bacillus licheniformis.

Food microbiology·2026
Same author

Spatiotemporal patterns and incidence trends of other infectious diarrhea in Qinghai Province, 2009-2023: a surveillance-based analysis.

Frontiers in public health·2026
Same author

Construction and validation of risk prediction model for glioblastoma associated with cancer stem cells and disulfidptosis.

Translational cancer research·2026

Related Experiment Video

Updated: Oct 13, 2025

Three-Dimensional Shape Modeling and Analysis of Brain Structures
05:33

Three-Dimensional Shape Modeling and Analysis of Brain Structures

Published on: November 14, 2019

7.3K

SAP-Net: A Simple and Robust 3D Point Cloud Registration Network Based on Local Shape Features.

Jinlong Li1, Yuntao Li1, Jiang Long1

  • 1School of Physical Science and Technology, Southwest Jiaotong University, Chengdu 610031, China.

Sensors (Basel, Switzerland)
|November 13, 2021
PubMed
Summary
This summary is machine-generated.

SAP-Net improves 3D point cloud registration by overcoming local optima and enhancing feature extraction. This novel approach offers superior performance and robustness compared to existing methods like ICP and CorsNet.

Keywords:
deep learningfeature extractionpoint cloudregistrationrobustness

More Related Videos

Four-Dimensional CT Analysis Using Sequential 3D-3D Registration
05:05

Four-Dimensional CT Analysis Using Sequential 3D-3D Registration

Published on: November 23, 2019

8.2K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

558

Related Experiment Videos

Last Updated: Oct 13, 2025

Three-Dimensional Shape Modeling and Analysis of Brain Structures
05:33

Three-Dimensional Shape Modeling and Analysis of Brain Structures

Published on: November 14, 2019

7.3K
Four-Dimensional CT Analysis Using Sequential 3D-3D Registration
05:05

Four-Dimensional CT Analysis Using Sequential 3D-3D Registration

Published on: November 23, 2019

8.2K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

558

Area of Science:

  • Computer Vision
  • 3D Data Processing
  • Machine Learning

Background:

  • Point cloud registration is crucial for 3D model reconstruction but traditional ICP algorithms are sensitive to initial positions and can get stuck in local optima.
  • Existing feature learning methods, such as PointNet, struggle with effective local feature extraction for registration.
  • There is a need for robust and accurate point cloud registration methods that are less dependent on initial conditions and better at capturing local geometric details.

Purpose of the Study:

  • To propose SAP-Net, an optimized network inspired by CorsNet and PointNet++, to address the limitations of traditional and existing deep learning-based point cloud registration methods.
  • To enhance the accuracy and robustness of point cloud registration by improving feature extraction and overcoming local optima.
  • To directly predict the rigid transformation parameters (rotation matrix and translation vector) for point cloud alignment.

Main Methods:

  • SAP-Net utilizes the set abstraction layer from PointNet++ for efficient feature extraction.
  • It integrates global features with the initial template point cloud to enrich contextual information.
  • PointNet serves as the final transform prediction layer to directly output the six degrees of freedom transformation parameters.

Main Results:

  • SAP-Net demonstrated superior performance over Iterative Closest Point (ICP) and CorsNet on the ModelNet40 dataset.
  • The proposed method achieved better results on both seen and unseen categories of point clouds.
  • Experiments confirmed SAP-Net's enhanced robustness in point cloud registration tasks.

Conclusions:

  • SAP-Net effectively addresses the limitations of traditional ICP and existing deep learning methods for point cloud registration.
  • The network architecture successfully improves feature extraction and avoids local optima, leading to more accurate alignments.
  • SAP-Net offers a robust and high-performing solution for 3D data model reconstruction.