Basic use of Open3D

Foreword

When processing point cloud data, the installation and getting started with the PCL library is more difficult. Intel's Open3D is excellent in performance and unpacking, it can read, process, and display various dot clouds, so this test is done.

Preparation

Install

Use Pychram under Window11 to configure the virtual environment of Python3.6 (PS: Open3D to blog writing date, temporarily only supports version 3.6-3.8), and setting of the virtual environment. Install Open3D, NUMPY, OpenCV (Optional).

Test data download

Loat the source code of Open3D on getHub, which contains the test data we need to be officially provided:

isl-org/Open3D: Open3D: A Modern Library for 3D Data Processing (github.com)

Test data is located in:

Open3D-master\\examples\\test_data\\

Preliminary use

Point cloud display

Simple use, display a colorful point cloud

import open3d as o3d
import numpy as np

print("Load a ply point cloud, print it, and render it")
#Read point cloud data
pcd = o3d.io.read_point_cloud("C:\\Document\\Open3D-master\\examples\\test_data\\fragment.ply")
print(pcd)
print(np.asarray(pcd.points))
#Show point cloud data
o3d.visualization.draw_geometries([pcd],
                                  zoom=0.3412,
                                  front=[0.4257, -0.2125, -0.8795],
                                  lookat=[2.6172, 2.0475, 1.532],
                                  up=[-0.0694, -0.9768, 0.2024])

The display effect is shown below:
Insert a picture description here

voxel sampling

Voxel sampling uses conventional virgin grids to create uniform sampling point clouds from the input point cloud. It is usually used for pre -processing steps for many cloud processing tasks. This algorithm is divided into two steps:

1. Points to be folded into Voxel.

2. Each occupied virus is exactly one point through all points within the average.

The code changed a little on the basis of just now:

import open3d as o3d
import numpy as np

print("Load a ply point cloud, print it, and render it")
pcd = o3d.io.read_point_cloud("C:\\Document\\Open3D-master\\examples\\test_data\\fragment.ply")
print("Downsample the point cloud with a voxel of 0.05")
downpcd = pcd.voxel_down_sample(voxel_size=0.05)
o3d.visualization.draw_geometries([downpcd],
                                  zoom=0.3412,
                                  front=[0.4257, -0.2125, -0.8795],
                                  lookat=[2.6172, 2.0475, 1.532],
                                  up=[-0.0694, -0.9768, 0.2024])

The dense point clouds in the previous step have become relatively sparse, and their uses need to be further learned.

The fixed -point method is estimated

Another basic operation of Dianyun is the dotted line estimate. Press the n key to check the point method. Bonded-and+can be used to control the length of the method.

import open3d as o3d
import numpy as np

print("Load a ply point cloud, print it, and render it")
pcd = o3d.io.read_point_cloud("C:\\Document\\Open3D-master\\examples\\test_data\\fragment.ply")
print("Downsample the point cloud with a voxel of 0.05")
downpcd = pcd.voxel_down_sample(voxel_size=0.05)
downpcd.estimate_normals(
    search_param=o3d.geometry.KDTreeSearchParamHybrid(radius=0.1, max_nn=30))
o3d.visualization.draw_geometries([downpcd],
                                  zoom=0.3412,
                                  front=[0.4257, -0.2125, -0.8795],
                                  lookat=[2.6172, 2.0475, 1.532],
                                  up=[-0.0694, -0.9768, 0.2024],
                                  point_show_normal=True)

-Effect

+Effect

PS: ESTIMATE_NORMALS calculates the method of each point. This function is used to find adjacent points, and uses a covariance analysis to calculate the main axis of the adjacent point.

This function uses instances of KDTREESEARCHPARAMHYBRID class as a parameter. Two key parameters Radius = 0.1 and max_nn = 30 Specify the search radius and the maximum nearest neighbor. Its search radius is 10cm, and only 30 neighbors are considered to save computing time.