Research on 3D Point Cloud Denoising Method Based on KNN Clustering
In order to solve the problem of random Gaussian noise in three-dimensional point clouds,a three-dimensional point cloud denoising method based on hierarchical K-Nearest Neighbor(KNN)clustering is proposed.Firstly,KNN is used to find the nearest K neighbors around each data point,and the random Gaussian noise of 3D point cloud data is preliminarily denoised according to the neighbor values,so as to reduce the influence of isolated points.Then,the distance-based clustering method is used to group the 3D point cloud data according to the distance,further de-noising the point cloud data,identifying the main cluster and eliminating the secondary cluster to improve the 3D point cloud de-noising accuracy.Experimental results show that the peak signal-to-noise ratio of KNN clustering algorithm is increased by 4～15dB,show-ing advantages in denoising effect and preserving point cloud details.It can not only better re-move three-dimensional point cloud noise,but also retain the geometric features and details of point cloud,and show higher performance and robustness in three-dimensional point cloud de-noising effect.
KNN clustering algorithmThree-dimensional point cloudRandom noisePeak sig-nal-to-noise ratio
NETL
NSTL
万方数据
