Indoor Scene Point Cloud Registration Algorithm Based on RGB-D Camera Calibration.

Area of Science:

• Computer Vision
• Robotics
• 3D Reconstruction

Background:

• RGB-depth (RGB-D) sensors are increasingly popular for 3D indoor scene reconstruction.
• Efficiently aligning multiple point clouds is crucial for accurate 3D modeling.

Purpose of the Study:

• To propose an automatic point cloud registration algorithm for 3D indoor scene reconstruction.
• To efficiently handle data from RGB-D sensors on fixed pan-tilt platforms.

Main Methods:

• Utilizes extrinsic parameters of RGB-D cameras from preset pan-tilt control points.
• Employs a computationally efficient global registration method based on offline calibrated extrinsic parameters.
• Includes an optional local registration method for refining alignment.

Main Results:

• The proposed algorithm demonstrates quality and computational efficiency in point cloud alignment.
• Experimental results validate the effectiveness compared to state-of-the-art methods.

Conclusions:

• The developed algorithm offers an efficient solution for 3D indoor scene reconstruction using RGB-D data.
• The method is suitable for fixed-position setups with pan-tilt platforms.