基于法向修正与位置滤波的散乱点云去噪算法

Denoising algorithm for scattered point clouds based on normal correction and position filtering

为解决传统点云去噪算法易造成模型的过光顺和局部失真问题,提出一种基于法向修正与位置滤波的散乱点云去噪算法。首先利用主成分分析(PCA)算法初估点云法向,对采样点的邻域点法向进行双边加权,利用加权后的邻域点法向之和估计原采样点的高精度法向;利用采样点与邻域点的平均距离及距离偏差在采样点法向上投影的标准差控制滤波参数的取值,使滤波参数能随采样点的局部几何特征自适应变化;最后将修正后的法向和参数自适应的高斯核函数结合,构建新的点云滤波模型。实验结果表明:修正后的法向偏差均方根为0.0624 rad,滤波后的点云数据的最小平均误差为0.0167 mm,最小均方根误差为0.4636μm,最大误差为0.0332 mm。该算法有效地避免了模型的过光顺和特征细节失真。

In order to solve the problem of over-smoothness and local distortion caused by traditional point cloud denoising algorithm,a scattershot point cloud denoising algorithm based on normal correction and position filtering is proposed.Firstly,the principal component analysis(PCA)algorithm is used to estimate the point cloud normal direction,then the neighborhood normal direction of the sampling point is weighted bilally,and the high precision normal direction of the original sampling point is estimated by the weighted sum of the neighborhood normal directions.The filtering parameters are controlled by the standard deviation of the average distance between the sampling points and the neighborhood points and the projection of the distance deviation on the sampling point method,so that the filtering parameters can change adaptively with the local geometric features of the sampling points.Finally,a new point cloud filter model is constructed by combining the modified normal with the parameter adaptive Gaussian kernel function.The experimental results show that the normal mean square root corrected method is 0.0624 rad for the deviation.The minimum average error of filtered point cloud data is 0.0167 mm.The minimum root mean square error is 0.4636μm,the maximum error is 0.0332 mm.The algorithm can effectively avoid over-smoothness and feature detail distortion of the model.