三维点云发现中值滤波

import open3d as o3d
import numpy as np
from scipy.spatial import cKDTree
from sklearn.decomposition import PCA
from concurrent.futures import ThreadPoolExecutor
def median_filter(points, k):
filtered_points = []
# 构建KD树
tree = cKDTree(points)

for i in range(len(points)):
# 在KD树中查询最近的k个邻近点
_, indices = tree.query([points[i]], k)
neighborhood = points[indices[0][1:]]

if len(neighborhood) > 0:
# 使用PCA计算邻域点集的主成分
pca = PCA(n_components=3)
pca.fit(neighborhood)
eigenvalues = pca.explained_variance_
eigenvectors = pca.components_

# 找到最小特征值对应的特征向量
min_eigenvalue_index = np.argmin(eigenvalues)
normal = eigenvectors[min_eigenvalue_index]

filtered_points.append(normal)

return filtered_points


point_cloud = o3d.io.read_point_cloud("bone3.ply")
o3d.visualization.draw_geometries([point_cloud])
points = np.array(point_cloud.points)
####进行法线滤波

radius = 5
k = 10
normal = median_filter(points, k)
np.savetxt('normal3.txt', normal, fmt='%.6f')