Incremental Structure from Motion for Small-Scale Scenes Based on Auxiliary Calibration.

Area of Science:

• Computer Vision
• Robotics
• 3D Reconstruction

Background:

• Sparse feature points limit incremental structure from motion (SfM) accuracy in small-scale scenes.
• Existing SfM methods struggle with stability and completeness in constrained environments.

Purpose of the Study:

• To propose an enhanced incremental SfM method for small-scale scenes.
• To improve feature point density and accuracy in sparse regions.
• To achieve more robust and precise 3D reconstructions.

Main Methods:

• Integration of an auxiliary calibration plate to increase feature points.
• Random generation of feature points in sparse areas.
• Utilizing pairwise polar geometric constraints for coarse feature matching.
• Filtering high-precision matching points using calibration plate geometric constraints.

Main Results:

• Demonstrated superior reconstruction completeness and accuracy compared to competing methods.
• Achieved average re-projection errors of 0.5245, 0.4151, and 0.4996 pixels in real-world experiments.
• Ensured robust pose estimation and precise 3D reconstructions.

Conclusions:

• The proposed SfM method effectively addresses the challenge of sparse feature points in small-scale scenes.
• The calibration plate integration significantly enhances 3D reconstruction quality.
• This approach offers a robust solution for accurate and stable SfM in practical applications.