A high precision 3D reconstruction method for bend tube axis based on binocular stereo vision

Area of Science:

• Mechanical Engineering
• Computer Vision
• Metrology

Background:

• Stereo-vision methods for bent tube axis measurement often use image contour centerlines, leading to reconstruction errors.
• These errors are exacerbated by increasing tube diameter, limiting accuracy in practical applications.

Purpose of the Study:

• To develop a novel stereo-vision approach for precise 3D reconstruction of bent tube axes.
• To overcome the limitations of existing methods, particularly concerning tube diameter variations.

Main Methods:

• Established a perspective projection model for arbitrary cross-sections of bent tubes.
• Developed a method to accurately locate projected axis points on image planes.
• Reconstructed 3D axis coordinates using binocular stereo vision.

Main Results:

• The proposed method effectively reduces reconstruction errors compared to classical approaches.
• Measurement accuracy was found to be practically independent of the tube's diameter.
• Validated through measurements on three bent tubes with varying diameters.

Conclusions:

• The novel perspective projection model and precise point localization significantly improve bent tube axis measurement accuracy.
• This method offers a robust solution for metrology applications involving bent tubes of diverse diameters.