Unorganized 3D Point Clouds Denoising and Sharpening

The increasing use of 3D model reconstruction in industry has raised demand for accurate and cost-effective point cloud data. In industrial automation, creating 3D point clouds is recognized for reverse engineering modeling, yet noise from depth sensors or LiDARs can impact accuracy. This paper introduces an adaptive scaling strategy using a Weighted Principal Component Analysis (WPCA) and an Improved Bilateral Filtering (IBF) to address denoising and sharpening challenges in unorganized 3D point clouds. The method proves robust against high noise levels, effectively preserving key features. Our research focuses on unorganized point clouds but is adaptable to organized ones. Validation involves examining near-surface and open-surface point cloud data, and the approach's efficacy is confirmed through three-dimensional model reconstruction with real-world data.