A Classification Denoising Algorithm Based on 3D Point Cloud Curvature

Abstract

In the process of denoising and smoothing 3D point cloud data, it often faces the problems of poor results required for denoising and difficulty in maintaining the local feature details of the point cloud. These problems directly affect the accuracy and computational complexity of subsequent 3D model reconstruction. Therefore, this study proposes a curvature-based classification algorithm for denoising 3D point cloud data, where the point cloud data is classified into smooth and sharp regions based on a set curvature threshold. Statistical filtering combined with radius filtering and Gaussian filtering are used to achieve the processing of different regions, respectively, which effectively achieves fast denoising of the point cloud model. The experimental results show that the proposed method not only effectively removes the noisy points but also preserves the local features of the point cloud model data by applying this method and a single denoising method to the self-collected three-dimensional point cloud data.