点云滤波：半径滤波python实现以及open3d实现

半径滤波与统计滤波器类似，根据空间点半径范围中临近点数量是否满足给定值来滤波。该滤波算法比统计滤波更加简单，计算速度更快。

在点云数据中以某点为中心画一个圆计算落在该圆中点的数量，当数量大于给定值时，则保留该点，数量小于给定值则剔除该点。此算法运行速度快，依序迭代留下的点一定是最密集的，但是圆的半径和圆内点的数目都需要人工指定。

（1）open3d实现半径滤波：

import open3d as o3d
import numpy as np
pcd = o3d.io.read_point_cloud('013205.pcd',remove_nan_points = True,remove_infinite_points = True)
print('原点云个数为：',np.array(pcd.points).shape[0])
o3d.visualization.draw_geometries([pcd],window_name = '源数据',width = 1080,height = 720,top = 30,left = 20)
cl,index = pcd.remove_radius_outlier(nb_points=10,radius=0.2)
new_pcd = pcd.select_by_index(index)
print('半径滤波后的点云个数为：',np.array(new_pcd.points).shape[0])
o3d.visualization.draw_geometries([new_pcd],window_name= '半径滤波',width = 1080,height = 720,top = 20,left = 30)

（2）python代码实现半径滤波

import open3d as o3d
import numpy as np
pcd = o3d.io.read_point_cloud('013205.pcd',remove_nan_points = True,remove_infinite_points = True)
print('原点云个数为：',np.array(pcd.points).shape[0])
o3d.visualization.draw_geometries([pcd],window_name = '源数据',width = 1080,height = 720,top = 30,left = 20)

def radius_outlier(cloud,nb_points = 10,radius = 0.2):
    #点云数据构造kd树
    pcd_tree = o3d.geometry.KDTreeFlann(cloud)
    #寻找每个点的邻近nb_points个数的点，并且计算他们之间的距离，如果距离大于radius，则舍弃该点，小于则保留
    di = []
    new_cloud = []
    for i in range(np.array(cloud.points).shape[0]):#遍历所有点
        [k,idx,_] = pcd_tree.search_knn_vector_3d(cloud.points[i],nb_points)
        #计算该点到每个点的欧式距离
        euc_distance = [ np.linalg.norm(np.array(cloud.points)[j] - np.array(cloud.points)[idx[0]])for j in np.array(idx)[1:]]
        is_less_than_radius = [j for j in euc_distance if j > radius ]
        if len(is_less_than_radius) == 0 :#所有距离都符合
            new_cloud.append(np.array(cloud.points)[i])
    new_pcd = o3d.geometry.PointCloud()
    points = o3d.utility.Vector3dVector(np.array(new_cloud))
    new_pcd.points = points
    o3d.io.write_point_cloud('radius_deal_points.pcd',new_pcd,True)
    o3d.visualization.draw_geometries([new_pcd])
radius_outlier(pcd,10,0.2)