An efficient outlier removal method for scattered point cloud data.

Area of Science:

• Computer Vision
• Geometric Data Processing
• 3D Scanning Technologies

Background:

• Scanned point cloud data (PCD) quality is critical for industrial applications and reverse engineering.
• Acquired PCD is often noisy, sparse, and temporarily incoherent, posing ill-posed processing challenges.
• Existing outlier removal methods may struggle with complex, non-isolated noise patterns.

Purpose of the Study:

• To develop a simple and effective method for removing noisy points from scanned PCD.
• To enhance the quality and reliability of point cloud data for downstream applications.
• To address limitations in current outlier removal techniques for complex datasets.

Main Methods:

• A novel method utilizing two geometrical characteristic constraints: local density and deviation from a local fitting plane.
• Local density analysis for preprocessing to remove sparse and isolated outliers.
• A local projection method to identify and remove non-isolated outliers based on plane deviation.

Main Results:

• Experimental results demonstrate successful removal of noisy points from various man-made objects.
• The method effectively handles complex outlier scenarios, improving data quality.
• Validation shows the robustness of the proposed geometrical constraints.

Conclusions:

• The presented method offers a robust and efficient solution for outlier removal in scanned PCD.
• The combined use of local density and plane deviation effectively addresses diverse outlier types.
• This approach significantly improves the quality of point cloud data for industrial and reverse engineering tasks.