Research on Point Cloud Filtering Algorithm Based on Kmeans++ Clustering and WLOP

In computer vision and 3D data processing, effective point cloud filtering is crucial for high-quality model reconstruction and analysis. Traditional methods like the weighted local optimal projection (WLOP) can reduce noise but often lead to over-smoothing, especially in complex or dense point clouds. This study introduces a novel point cloud filtering technique that integrates Kmeans++ clustering with an improved WLOA methodology. By pre-classifying the point cloud data with Kmeans++, the method enhances filtering efficiency and preserves more structural details. Experimental results demonstrate that this approach significantly outperforms traditional WLOP in terms of filtering accuracy and edge detail retention, particularly in large-scale or noisy environments. Additionally, this research assesses the impact of different clustering techniques on filtering performance, highlighting the effectiveness of Kmeans++ clustering within this framework. These findings advance point cloud processing technology and offer new insights and tools for further research and practical applications in the field.