嗨，针对你说的从非规则轮廓点云（每个z值对应一组x、y点，而且x、y不像标准地形图那样有唯一性）生成曲面或网格的需求，结合你提到的双球轮廓示例——也就是一个大球的赤道轮廓环，加上小球与大球相交形成的上层轮廓环的点云，我整理了几个实用的Python实现思路，帮你更清晰地可视化这些点的形状：

首先先补全你提到的点云生成代码，方便后续测试：

import numpy as np
import pandas as pd
import plotly.graph_objects as go
import plotly.express as px

# 生成大球的z=0层赤道轮廓点
theta = np.linspace(0, 2*np.pi, 100)
x1 = 2*np.cos(theta)
y1 = 2*np.sin(theta)
z1 = np.zeros_like(x1)

# 生成小球与大球的交线轮廓点（联立两球方程得到z=2的交线）
phi = np.linspace(0, 2*np.pi, 100)
x2 = np.cos(phi)
y2 = np.sin(phi)
z2 = np.sqrt(4 - x2**2 - y2**2)

# 合并所有点并转为DataFrame
x = np.concatenate([x1, x2])
y = np.concatenate([y1, y2])
z = np.concatenate([z1, z2])
points = pd.DataFrame({"x": x, "y": y, "z": z})

方法1：基于凸包的分层网格绘制 #

因为你的点是按z值分层的闭合轮廓，我们可以对每个z层的点计算2D凸包，再把这些凸包面组合成3D网格：

from scipy.spatial import ConvexHull

# 按z值分组处理每个轮廓层
groups = points.groupby("z")
vertices = []
faces = []
current_idx = 0

for z_val, group in groups:
    # 获取当前层的三维点
    layer_pts = group[["x", "y", "z"]].values
    vertices.extend(layer_pts.tolist())
    
    # 计算2D凸包（仅用x、y坐标）
    hull = ConvexHull(layer_pts[:, :2])
    # 将凸包的多边形面拆分为三角形（适配Plotly的Mesh3d格式）
    for simplex in hull.simplices:
        faces.append([current_idx + simplex[0], current_idx + simplex[1], current_idx + simplex[2]])
    current_idx += len(layer_pts)

# 绘制网格+原始点云对比
fig = go.Figure(data=[go.Mesh3d(
    x=[v[0] for v in vertices],
    y=[v[1] for v in vertices],
    z=[v[2] for v in vertices],
    i=[f[0] for f in faces],
    j=[f[1] for f in faces],
    k=[f[2] for f in faces],
    opacity=0.5,
    color='lightblue'
)])

fig.add_trace(go.Scatter3d(
    x=points["x"], y=points["y"], z=points["z"],
    mode='markers', marker=dict(size=3, color='crimson')
))

fig.update_layout(scene=dict(aspectmode='data'))
fig.show()

方法2：3D Delaunay剖分生成表面网格 #

如果你的点云是连续的轮廓集合，可以用3D Delaunay三角剖分，再筛选出表面的三角面来生成网格：

from scipy.spatial import Delaunay, ConvexHull

# 对所有点进行3D Delaunay剖分
tri = Delaunay(points[["x", "y", "z"]].values)
# 通过凸包筛选出表面的三角面
hull = ConvexHull(points[["x", "y", "z"]].values)
surface_faces = tri.simplices[hull.simplices]

# 可视化结果
fig = go.Figure(data=[go.Mesh3d(
    x=points["x"], y=points["y"], z=points["z"],
    i=[f[0] for f in surface_faces],
    j=[f[1] for f in surface_faces],
    k=[f[2] for f in surface_faces],
    opacity=0.6,
    color='orange'
)])

fig.add_trace(go.Scatter3d(
    x=points["x"], y=points["y"], z=points["z"],
    mode='markers', marker=dict(size=2, color='darkblue')
))
fig.show()

方法3：用Open3D做专业的泊松表面重建 #

如果你的点云更复杂（比如非闭合、有噪声），Open3D的泊松重建算法能生成更平滑的曲面：

import open3d as o3d

# 转换点云为Open3D格式
pcd = o3d.geometry.PointCloud()
pcd.points = o3d.utility.Vector3dVector(points[["x", "y", "z"]].values)

# 估计点云的法向量（泊松重建必需）
pcd.estimate_normals(search_param=o3d.geometry.KDTreeSearchParamHybrid(radius=0.1, max_nn=30))

# 执行泊松重建
mesh, densities = o3d.geometry.TriangleMesh.create_from_point_cloud_poisson(pcd, depth=9)

# 可选：过滤低密度的冗余面，让网格更干净
vertices_to_remove = densities < np.quantile(densities, 0.1)
mesh.remove_vertices_by_mask(vertices_to_remove)

# 打开Open3D可视化窗口
o3d.visualization.draw_geometries([mesh])

备选方案：插值生成规则曲面 #

如果只是想快速展示轮廓的形状，也可以先对每个z层的点做插值，生成规则网格后绘制曲面：

from scipy.interpolate import griddata

# 生成规则的x、y网格
xi = np.linspace(points["x"].min(), points["x"].max(), 50)
yi = np.linspace(points["y"].min(), points["y"].max(), 50)
xi, yi = np.meshgrid(xi, yi)

fig = go.Figure()
# 对每个z层的点进行插值并绘制曲面
for z_val in points["z"].unique():
    subset = points[points["z"] == z_val]
    zi = griddata((subset["x"], subset["y"]), subset["z"], (xi, yi), method='cubic')
    fig.add_trace(go.Surface(x=xi, y=yi, z=zi, opacity=0.4))

# 叠加原始点云
fig.add_trace(go.Scatter3d(
    x=points["x"], y=points["y"], z=points["z"],
    mode='markers', marker=dict(size=3, color='purple')
))
fig.show()

备注：内容来源于stack exchange，提问作者Lluis