Multi-level k -nearest neighbors algorithm for direct point cloud-based engineering analysis

Area of Science:

• Computational geometry
• Computer-aided engineering
• Geometric modeling

Background:

• Point clouds are vital for geometric modeling in science and engineering due to advanced scanning technologies.
• The unstructured nature of point clouds poses challenges for direct engineering analysis, often necessitating mesh reconstruction.
• Existing mesh reconstruction methods, including k-nearest neighbors (kNN), require manual parameter tuning for complex or noisy data.

Purpose of the Study:

• To develop an automated and robust method for processing point cloud data without manual intervention.
• To enhance the accuracy of geometric processing and enable direct analysis of raw point clouds in engineering applications.
• To overcome limitations of traditional kNN algorithms in handling varying point cloud densities and topological complexities.

Main Methods:

• Introduction of a novel multi-level k-nearest neighbors (M-kNN) approach.
• Iterative expansion of local neighborhoods to determine surface connectivity from point cloud data.
• Application of M-kNN for point cloud resampling and geometry processing, particularly for complex structures.

Main Results:

• M-kNN demonstrates improved point cloud resampling and more accurate geometry processing, especially for intricate geometries.
• The method effectively handles noisy, low-density, and topologically ambiguous point clouds without manual tuning.
• Validation on both synthetic and real-world datasets confirms the efficacy of the M-kNN approach.

Conclusions:

• The M-kNN approach offers a significant advancement in processing point cloud data for engineering analysis.
• It enables the direct utilization of raw point clouds, eliminating the need for mesh reconstruction and manual parameter adjustments.
• This method enhances the flexibility and accuracy of geometric modeling and analysis in various scientific and engineering fields.