实时向点云添加新点 - Open3D

Question

我正在使用 Open3D 在 Python 中可视化点云。本质上，我想做的是以编程方式向点云添加另一个点，然后实时渲染它。

这是我到目前为止所拥有的。我找不到任何解决方案。

在下面的代码中，我展示了一种可能的解决方案，但它并不有效。第一个窗口关闭后，点将被添加并打开一个新窗口。这不是我想要的。我希望它能够动态显示新点，而无需再次关闭和打开。以及创建一个新变量的事实，我认为在处理越来越大的数据集时可能会出现问题

import open3d as o3d
import numpy as np

#Create two random points
randomPoints = np.random.rand(2, 3)

pointSet = o3d.geometry.PointCloud()

pointSet.points = o3d.utility.Vector3dVector(randomPoints)

#Visualize the two random points
o3d.visualization.draw_geometries([pointSet])

#Here I want to add more points to the pointSet
#This solution does not work effective

#Create another random set
p1 = np.random.rand(3, 3)

p2 = np.concatenate((pointSet.points, p1), axis=0)

pointSet2 = o3d.geometry.PointCloud()

pointSet2.points = o3d.utility.Vector3dVector(p2)

o3d.visualization.draw_geometries([pointSet2])

有没有可能解决这个问题？

如果没有，我还可以查看哪些其他库能够实时渲染新的传入点。

Answer 1

PointCloud通过扩展新坐标，可以将新点添加到 a 中并以交互方式可视化PointCloud.points。

\n

这是因为当我们使用 numpy 数组时，我们需要创建一个Vector3dVector实例来实现方便的方法extend。来自文档：

\n

\n

扩展(* args , ** kwargs )

\n

重载功能。

\n

\n
1. 扩展（自身：open3d.cpu.pybind.utility.Vector3dVector，L：\ nopen3d.cpu.pybind.utility.Vector3dVector） - >无
\n

\n

通过附加给定列表中的所有项目来扩展列表

\n

\n
2. 扩展（自我：open3d.cpu.pybind.utility.Vector3dVector，L：可迭代）->\n无
\n

\n

通过附加给定列表中的所有项目来扩展列表

\n

\n

所以我们可以使用不同的对象实例，例如，ndarrays等等。Vector3dVectorlists

\n

一个玩具示例及其结果：

\n

\n

import open3d as o3d\nimport numpy as np\nimport time\n\n\n# create visualizer and window.\nvis = o3d.visualization.Visualizer()\nvis.create_window(height=480, width=640)\n\n# initialize pointcloud instance.\npcd = o3d.geometry.PointCloud()\n# *optionally* add initial points\npoints = np.random.rand(10, 3)\npcd.points = o3d.utility.Vector3dVector(points)\n\n# include it in the visualizer before non-blocking visualization.\nvis.add_geometry(pcd)\n\n# to add new points each dt secs.\ndt = 0.01\n# number of points that will be added\nn_new = 10\n\nprevious_t = time.time()\n\n# run non-blocking visualization. \n# To exit, press \'q\' or click the \'x\' of the window.\nkeep_running = True\nwhile keep_running:\n    \n    if time.time() - previous_t > dt:\n        # Options (uncomment each to try them out):\n        # 1) extend with ndarrays.\n        pcd.points.extend(np.random.rand(n_new, 3))\n        \n        # 2) extend with Vector3dVector instances.\n        # pcd.points.extend(\n        #     o3d.utility.Vector3dVector(np.random.rand(n_new, 3)))\n        \n        # 3) other iterables, e.g\n        # pcd.points.extend(np.random.rand(n_new, 3).tolist())\n        \n        vis.update_geometry(pcd)\n        previous_t = time.time()\n\n    keep_running = vis.poll_events()\n    vis.update_renderer()\n\nvis.destroy_window()\n

\n\n

为什么不创建更新的几何图形并删除旧的几何图形呢？

\n

为了完整起见，其他（我认为不是更好）替代方法可以包括以下步骤：

\n

\n
1. 删除当前的PointCloud
\n
2. 连接OP问题中的新点
\n
3. 创建新的Pointcloud并将其添加到可视化工具中。
\n

\n

这会产生更差的性能，并且几乎不允许与可视化进行交互。

\n

为了了解这一点，让我们看一下以下具有相同设置的比较（代码如下）。两个版本运行时间相同（约 10 秒）。

\n

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

使用extend	删除和创建PointCloud
\xe2\x9c\x94\xef\xb8\x8f 允许交互	\xe2\x9d\x8c 交互困难
平均执行时间（用于添加点）：0.590 ms	平均执行时间（用于添加点）：1.550 ms

\n

\n

重现代码：

\n

import open3d as o3d\nimport numpy as np\nimport time\n\n# Global settings.\ndt = 3e-2 # to add new points each dt secs.\nt_total = 10 # total time to run this script.\nn_new = 10 # number of points that will be added each iteration.\n\n#---\n# 1st, using extend. Run non-blocking visualization.\n\n# create visualizer and window.\nvis = o3d.visualization.Visualizer()\nvis.create_window(height=480, width=640)\n\n# initialize pointcloud instance.\npcd = o3d.geometry.PointCloud()\n# *optionally* add initial points\npoints = np.random.rand(10, 3)\npcd.points = o3d.utility.Vector3dVector(points)\n# include it in the visualizer before non-blocking visualization.\nvis.add_geometry(pcd)\n\nexec_times = []\n\ncurrent_t = time.time()\nt0 = current_t\n\nwhile current_t - t0 < t_total:\n\n    previous_t = time.time()\n\n    while current_t - previous_t < dt:\n        s = time.time()\n\n        # Options (uncomment each to try it out):\n        # 1) extend with ndarrays.\n        pcd.points.extend(np.random.rand(n_new, 3))\n\n        # 2) extend with Vector3dVector instances.\n        # pcd.points.extend(\n        #     o3d.utility.Vector3dVector(np.random.rand(n_new, 3)))\n\n        # 3) other iterables, e.g\n        # pcd.points.extend(np.random.rand(n_new, 3).tolist())\n\n        vis.update_geometry(pcd)\n\n        current_t = time.time()\n        exec_times.append(time.time() - s)\n\n    vis.poll_events()\n    vis.update_renderer()\n\nprint(f"Using extend\\t\\t\\t# Points: {len(pcd.points)},\\n"\n      f"\\t\\t\\t\\t\\t\\tMean execution time:{np.mean(exec_times):.5f}")\n\nvis.destroy_window()\n\n# ---\n# 2nd, using remove + create + add PointCloud. Run non-blocking visualization.\n\n# create visualizer and window.\nvis = o3d.visualization.Visualizer()\nvis.create_window(height=480, width=640)\n\n# initialize pointcloud instance.\npcd = o3d.geometry.PointCloud()\npoints = np.random.rand(10, 3)\npcd.points = o3d.utility.Vector3dVector(points)\nvis.add_geometry(pcd)\n\nexec_times = []\n\ncurrent_t = time.time()\nt0 = current_t\nprevious_t = current_t\n\nwhile current_t - t0 < t_total:\n\n    previous_t = time.time()\n\n    while current_t - previous_t < dt:\n        s = time.time()\n\n        # remove, create and add new geometry.\n        vis.remove_geometry(pcd)\n        pcd = o3d.geometry.PointCloud()\n        points = np.concatenate((points, np.random.rand(n_new, 3)))\n        pcd.points = o3d.utility.Vector3dVector(points)\n        vis.add_geometry(pcd)\n\n        current_t = time.time()\n        exec_times.append(time.time() - s)\n\n    current_t = time.time()\n\n    vis.poll_events()\n    vis.update_renderer()\n\nprint(f"Without using extend\\t# Points: {len(pcd.points)},\\n"\n      f"\\t\\t\\t\\t\\t\\tMean execution time:{np.mean(exec_times):.5f}")\n\nvis.destroy_window()\n

\n