Improved Method for 3D Point Cloud Reconstruction Based on Filtering

Point cloud denoising is essential for reconstructing high-quality surfaces with details, and aims to eliminate noise and outliers from the 3D scanning process. The challenge for denoising algorithms is to reduce noise and preserve sharp features. However, anomalous feature point clouds (especially high-density noisy point clouds) that survive the point cloud data processing can significantly discount the reconstruction quality. To address the above problems, we propose a new algorithm to eliminate these erroneous reconstructed 3D points. First, the 3D structure of the object is gradually recovered based on two point clouds using a matching algorithm. Then, we apply an improved matched point pair mapping filtering algorithm to the above point clouds to remove the erroneous 3D points and retrieve new matches. The experimental results show that our method can effectively remove the multi-scale noise from the point cloud model while better preserving its geometric features.