Prism-based self-calibration virtual cameras to generate rotating view fields for 6DOF pose measurement

Area of Science:

• Robotics and Computer Vision
• Metrology and Measurement Science

Background:

• Accurate pose measurement is critical for autonomous systems and manufacturing.
• Traditional methods often require complex camera calibration, limiting their use in constrained spaces.

Purpose of the Study:

• To propose a novel self-calibrating virtual camera system for six degrees of freedom (6DOF) pose determination.
• To enable accurate 6DOF measurement without explicit camera calibration in spatially constrained environments.

Main Methods:

• Development of a rotating-prism-embedded camera for self-calibration.
• Utilizing four control points for camera self-calibration, including focal length and distortion coefficient estimation via maximum likelihood estimation.
• Employing multi-view prior information fusion, rotation averaging, and translation averaging for accurate 6DOF recovery.

Main Results:

• The proposed method achieves higher measurement accuracy, stability, and efficiency compared to mainstream uncalibrated techniques.
• Demonstrated effectiveness even with significant noise and a limited number of control points.
• Successfully corrected image distortions from lenses and prisms, enabling 6DOF extraction with an uncalibrated camera.

Conclusions:

• The novel rotating-prism camera system offers a robust solution for uncalibrated 6DOF pose measurement.
• This approach is particularly advantageous for applications in spatially constrained scenes, such as autonomous navigation and smart manufacturing.