A Novel Method for Extrinsic Calibration of Multiple RGB-D Cameras Using Descriptor-Based Patterns

Area of Science:

• Computer Vision
• Robotics
• 3D Reconstruction

Background:

• Multi-camera systems are crucial for expanding field of view in applications like indoor 3D mapping and robot navigation.
• Calibrating RGB-D cameras with limited overlapping fields of view presents a significant challenge.

Purpose of the Study:

• To propose a novel method for estimating relative poses between RGB-D cameras with minimal overlapping fields of view.
• To enable wider field of view coverage for enhanced 3D mapping and robot navigation.

Main Methods:

• Utilizes descriptor-based patterns to identify well-matched 2D keypoints across cameras with minimal overlap.
• Integrates matched 2D keypoints with their corresponding depth values to construct 3D matched keypoints.
• Employs these 3D keypoints for the calibration of multiple RGB-D cameras.

Main Results:

• Demonstrates accurate estimation of relative camera poses.
• Achieves efficient calibration, suitable for real-time applications.
• Validates the approach through experimental results.

Conclusions:

• The proposed method effectively calibrates RGB-D cameras even with minimal visual overlap.
• This technique enhances the feasibility of using multiple RGB-D cameras for advanced robotic and mapping tasks.
• The approach offers a robust solution for challenging multi-camera calibration scenarios.