【小白深度教程 1.27】手把手教你使用 Open3D（10）点云地面拟合与地面分割（Python 代码）

在这篇文章中，我们将学习如何使用 Open3D 对 自动驾驶 中获取的激光雷达数据（点云），进行地面拟合与分割。

该代码主要用于从 .bin 文件加载点云数据，进行地面分割，并对结果进行可视化处理。主要步骤包括：加载点云数据、计算法向量、基于主成分分析 (PCA) 进行地面分割、并可视化处理结果。

可视化效果：

点云地面分割结果：

Python 源码：

#!/usr/bin/env python3

import numpy as np
import open3d as o3d
import copy

def pcd_ground_seg_pca(scan, th=0.80, z_offset=-1.1):
    """ Perform PCA over PointCloud to segment ground. """
    pcd = copy.deepcopy(scan)
    _, covariance = pcd.compute_mean_and_covariance()
    eigen_vectors = np.linalg.eig(covariance)[1]
    k = eigen_vectors.T[2]

    # magnitude of projecting each face normal to the z axis
    normals = np.asarray(scan.normals)
    points = np.asarray(scan.points)
    mag = np.linalg.norm(np.dot(normals, k).reshape(-1, 1), axis=1)
    ground = pcd.select_by_index(np.where((mag >= th) & (points[:, 2] < z_offset))[0])
    rest = pcd.select_by_index(np.where((mag >= th) & (points[:, 2] < z_offset))[0], invert=True)

    # Also remove the faces that are looking downwards
    up_normals = np.asarray(ground.normals)
    orientation = np.dot(up_normals, k)
    ground = ground.select_by_index(np.where(orientation > 0.0)[0])

    ground.paint_uniform_color([1.0, 0.0, 0.0])
    rest.paint_uniform_color([0.0, 0.0, 1.0])

    return ground, rest

def load_bin_file(file_path):
    """ Load a .bin point cloud file. """
    # Load the point cloud data assuming each point is [x, y, z] format
    point_cloud = np.fromfile(file_path, dtype=np.float32).reshape(-1, 4)[:, :3]  # Assuming (x, y, z, intensity)
    return point_cloud

def main():
    # Load the .bin file
    file_path = '0000000000.bin'  # Replace with your .bin file path
    points = load_bin_file(file_path)

    # Create an Open3D point cloud object
    pcd = o3d.geometry.PointCloud()
    pcd.points = o3d.utility.Vector3dVector(points)

    # Estimate normals (required for PCA)
    pcd.estimate_normals(search_param=o3d.geometry.KDTreeSearchParamHybrid(radius=1.0, max_nn=30))
    pcd.orient_normals_consistent_tangent_plane(100)

    # Visualize the original point cloud
    o3d.visualization.draw_geometries([pcd], window_name='Original Point Cloud')

    # Apply ground segmentation
    ground, rest = pcd_ground_seg_pca(pcd)

    # Visualize the segmented results
    o3d.visualization.draw_geometries([ground, rest], window_name='Segmented Point Cloud')

if __name__ == "__main__":
    main()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

分割前后对比：

原始点云，其中颜色用来标识高度：

分割结果：

示例文件下载：

通过网盘分享的文件：exp127
链接: https://pan.baidu.com/s/1OwUVyV4MMGtV4tCs4ikxmA 提取码: set8