Related Experiment Video
Updated: Jun 26, 2026
05:05
Four-Dimensional CT Analysis Using Sequential 3D-3D Registration
Published on: November 23, 2019
A Robust 3D Registration Method via Simultaneous Inlier Identification and Model Estimation
Xianyun Qian1, Fei Wen1, Peilin Liu1
1School of Integrated Circuits, Shanghai Jiao Tong University, Shanghai 200240, China.
Journal of Imaging
|June 25, 2026
Summary
This study introduces Simultaneous Inlier Identification and Model Estimation (SIME), a novel framework for robust 3D registration. SIME effectively handles noise and outliers, improving accuracy in computer vision and robotics applications.
Area of Science:
- Computer Vision
- Robotics
- Geometric Perception
Background:
- Robust 3D registration is crucial for estimating transformations between 3D data, but faces challenges from noise and outliers.
- Current methods like Maximum Consensus (MC) and M-estimators have limitations in jointly handling inlier identification and model estimation.
- A unified framework is needed for efficient and accurate 3D registration under challenging conditions.
Purpose of the Study:
- To introduce a unified framework for simultaneous inlier identification and model estimation (SIME) for robust 3D registration.
- To develop efficient algorithms for solving the proposed SIME formulation.
- To evaluate the performance of SIME against existing methods in various 3D registration scenarios.
Main Methods:
- Proposed a unified truncated-loss based formulation for Simultaneous Inlier Identification and Model Estimation (SIME).
- Developed an Alternating Minimization (AM) algorithm and an AM with Semidefinite Relaxation (AM-R) to solve the non-convex SIME problem.
- Applied the framework to 3D rotation search and rigid point-set registration using quaternion representations.
Main Results:
- SIME achieves lower fitting residuals than MC methods by incorporating residual magnitudes into inlier selection.
- The proposed AM and AM-R algorithms effectively solve the SIME formulation, handling binary inlier variables.
- Experimental results show SIME outperforms strong baselines, especially in high noise and extreme outlier scenarios (up to 95% outliers).
- On the 3DMatch dataset, SIME (AM) achieved a 91.0% mean registration success rate.
Conclusions:
- SIME offers a unified and efficient approach for simultaneous inlier identification and model estimation in 3D registration.
- The proposed methods demonstrate superior performance and robustness compared to existing techniques, particularly in challenging environments.
- SIME shows significant potential for reliable 3D registration in practical computer vision, robotics, and geometric perception applications.
Related Concept Videos
01:24One-Compartment Open Model: Wagner-Nelson and Loo Riegelman Method for ka Estimation
This lesson introduces two critical methods in pharmacokinetics, the Wagner-Nelson and Loo-Riegelman methods, used for estimating the absorption rate constant (ka) for drugs administered via non-intravenous routes. The Wagner-Nelson method relates ka to the plasma concentration derived from the slope of a semilog percent unabsorbed time plot. However, it is limited to drugs with one-compartment kinetics and can be impacted by factors like gastrointestinal motility or enzymatic degradation.
On...
On...
01:12Modeling and Similitude
Scaled modeling is a fundamental technique in engineering, enabling the study of large and complex systems by creating smaller, manageable replicas that recreate critical characteristics of the original. In hydrology and civil infrastructure, for example, scaled models of dams help analyze water flow, turbulence, and pressure. This method allows for accurate predictions of real-world behavior within a controlled environment, significantly reducing the cost and time involved in full-scale...