Globally optimal camera-and-rotation-sensor calibration with a branch-and-bound algorithm.

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Accurate estimation of relative poses between sensors and cameras is crucial for many applications.
• Existing methods may not guarantee global optimality or handle various error norms effectively.

Purpose of the Study:

• To introduce a novel, globally optimal algorithm for computing the rotational displacement between a rotation sensor and a rigidly attached camera.
• To provide a robust method applicable to different error metrics (L1, L2, L-infinity norms).

Main Methods:

• Development of a branch-and-bound algorithm to find the globally optimal solution.
• Derivation of a bounding inequality and feasibility problem for efficient rotation space search.
• Minimization of geometrically meaningful error using L1, L2, or L-infinity norms.

Main Results:

• The proposed algorithm successfully finds the globally optimal rotational displacement.
• Experimental validation with both synthetic and real-world data demonstrates the algorithm's efficacy.
• The method is shown to be effective across different error norm choices.

Conclusions:

• The branch-and-bound algorithm provides a globally optimal solution for sensor-camera rotational displacement.
• This method offers a robust and efficient approach for pose estimation problems.
• The algorithm's performance is validated, suggesting its practical applicability in robotics and computer vision.