Open3d去噪算法remove_radius_outlier and remove_statistical_outlier

半径滤波

在给定阈值参数nb_points后，遍历点云所有点，对于点云中任意一点，设其半径radius内有K个点，当K< nb_points时，即可识别该点为噪声点，并去除该点。

pc,idx=pcd.remove_radius_outlier(nb_points, radius, print_progress=False)

nb_points：邻域球内的最少点个数，小于该个数为噪声点
radius： 邻域半径大小
pc：去噪后的点云
idx：去噪保留的点索引

import open3d as o3d
import numpy as np


def display_inlier_outlier(cloud, ind):
    inlier_cloud = cloud.select_by_index(ind)
    outlier_cloud = cloud.select_by_index(ind, invert=True)

    print("Showing outliers (red) and inliers (gray): ")
    outlier_cloud.paint_uniform_color([1, 0, 0])
    inlier_cloud.paint_uniform_color([0.8, 0.8, 0.8])
    o3d.visualization.draw([inlier_cloud, outlier_cloud])


if __name__ == "__main__":
    ptcloud_data = o3d.data.PLYPointCloud()

    print("Load a ply point cloud, print it, and render it")
    pcd = o3d.io.read_point_cloud(ptcloud_data.path)
    R = pcd.get_rotation_matrix_from_xyz((np.pi, 0, 0))
    pcd.rotate(R, center=(0, 0, 0))
    o3d.visualization.draw([pcd])

    print("Downsample the point cloud with a voxel of 0.02")
    voxel_down_pcd = pcd.voxel_down_sample(voxel_size=0.02)
    o3d.visualization.draw([voxel_down_pcd])

    print("Radius oulier removal")
    cl, ind = voxel_down_pcd.remove_radius_outlier(nb_points=16, radius=0.05)
    display_inlier_outlier(voxel_down_pcd, ind)

统计滤波

当判断点的nb_neighbors近邻平均距离大于【平均距离+std_ratio*σ】，即判定为噪声点，σ为标准差
平均距离：遍历点云中的每个点到任意点距离的平均值。

pc,idx=pcd.remove_statistical_outlier( nb_neighbors, std_ratio, print_progress=False)

参数：
nb_neighbors：指nb_neighbors个最近点，我们需要计算距离该点nb_neighbors个最近点的平均距离
std_ratio：标准差的倍数
返回值：
pc： 去噪后的点云
idx： 去噪保留的点索引

import open3d as o3d
import numpy as np


def display_inlier_outlier(cloud, ind):
    inlier_cloud = cloud.select_by_index(ind)
    outlier_cloud = cloud.select_by_index(ind, invert=True)

    print("Showing outliers (red) and inliers (gray): ")
    outlier_cloud.paint_uniform_color([1, 0, 0])
    inlier_cloud.paint_uniform_color([0.8, 0.8, 0.8])
    o3d.visualization.draw([inlier_cloud, outlier_cloud])


if __name__ == "__main__":
    ptcloud_data = o3d.data.PLYPointCloud()

    print("Load a ply point cloud, print it, and render it")
    pcd = o3d.io.read_point_cloud(ptcloud_data.path)
    R = pcd.get_rotation_matrix_from_xyz((np.pi, 0, 0))
    pcd.rotate(R, center=(0, 0, 0))
    o3d.visualization.draw([pcd])

    print("Downsample the point cloud with a voxel of 0.02")
    voxel_down_pcd = pcd.voxel_down_sample(voxel_size=0.02)
    o3d.visualization.draw([voxel_down_pcd])

    print("Statistical oulier removal")
    cl, ind = voxel_down_pcd.remove_statistical_outlier(nb_neighbors=20,
                                                        std_ratio=2.0)
    display_inlier_outlier(voxel_down_pcd, ind)