Robust Estimation of Absolute Camera Pose via Intersection Constraint and Flow Consensus.

Area of Science:

• Computer Vision
• Robotics
• Geometric Computer Vision

Background:

• Accurate camera pose estimation is crucial for many applications.
• Outliers in 3D-to-2D correspondences significantly degrade pose estimation accuracy.
• Existing outlier removal methods have limitations in applicability and robustness.

Purpose of the Study:

• To develop a general and accurate outlier removal strategy for camera pose estimation.
• To create a method that is robust to high outlier ratios and applicable to various data types.
• To enhance existing pose estimation techniques by mitigating the impact of outliers.

Main Methods:

• A nested outlier removal strategy with outer and inner modules.
• Outer module utilizes an intersection constraint where projection rays/planes of inliers meet at the camera center.
• Inner module employs flow consensus on 2D displacements or 3D arcs for inlier probability refinement.

Main Results:

• The proposed strategy effectively removes outliers from point and line correspondences.
• It demonstrates reliable and efficient performance even with high outlier ratios.
• The method shows improved accuracy and robustness compared to state-of-the-art approaches on synthetic and real-world data.

Conclusions:

• The developed outlier removal strategy offers a general and accurate solution for robust camera pose estimation.
• Its nested structure and novel constraints enable superior performance over existing methods.
• This approach significantly advances the field of camera pose estimation in the presence of noisy data.