Motion-Based Extrinsic Sensor-to-Sensor Calibration: Effect of Reference Frame Selection for New and Existing...

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Sensor-to-sensor extrinsic calibration is vital for multi-sensor systems.
• Motion-based hand-eye calibration is a common approach.
• Sensor trajectories often contain noise, affecting calibration accuracy.

Purpose of the Study:

• To investigate the impact of reference frame selection on sensor-to-sensor extrinsic calibration.
• To identify optimal reference frame selection strategies under various noise conditions.
• To compare proposed cost functions against state-of-the-art methods.

Main Methods:

• Simulations with varying noise levels were performed.
• Real-world data from the KITTI dataset was used for validation.
• Four state-of-the-art methods and two novel cost functions were evaluated.

Main Results:

• One proposed cost function with outlier rejection significantly improved performance with outliers.
• Reference frame selection yielded greater performance gains than algorithm selection.
• Commonly used reference frame selection methods in literature were found to be suboptimal.

Conclusions:

• Careful reference frame selection is critical for robust sensor calibration.
• Outlier rejection in cost functions enhances calibration accuracy.
• Relative error metrics are unreliable for assessing calibration performance.