Robust Line Segment Matching for Space-Based Stereo Vision via Multi-Constraint Global Optimization.

Area of Science:

• Computer Vision
• Robotics
• Photogrammetry

Background:

• Line segment matching is crucial for stereo vision, but challenging in space applications due to weak textures and illumination variations.
• Existing methods struggle with accuracy and robustness in these demanding environments.

Purpose of the Study:

• To develop a robust and accurate line segment matching framework for stereo vision.
• To address the specific challenges of space-based applications.

Main Methods:

• A multi-constraint progressive matching framework integrating epipolar geometry, coplanarity, local homography, angular consistency, and distance-ratio invariance.
• A unified cost matrix for quantitative geometric residual encoding and candidate evaluation.
• Hungarian algorithm optimization for global consistency and mismatch suppression under one-to-one constraints.

Main Results:

• The proposed method demonstrates superior accuracy and robustness compared to conventional approaches.
• Consistent performance across diverse stereo image pairs, validated by quantitative and qualitative evaluations.
• Improved computational efficiency for line matching.

Conclusions:

• The multi-constraint framework provides reliable line correspondences for stereo vision.
• The method is highly effective for challenging space-based applications.
• Offers a significant advancement in robust and efficient line segment matching.