OpenCV  5.0.0-pre
Open Source Computer Vision
Loading...
Searching...
No Matches
Point cloud visualisation
Original author Dmitrii Klepikov
Compatibility OpenCV >= 5.0

Goal

In this tutorial you will:

  • Load and save point cloud data
  • Visualise your data

Requirements

For visualisations you need to compile OpenCV library with OpenGL support. For this you should set WITH_OPENGL flag ON in CMake while building OpenCV from source.

Practice

Loading and saving of point cloud can be done using cv::loadPointCloud and cv::savePointCloud accordingly.

Currently supported formats are:

  • .OBJ (supported keys are v(which is responsible for point position), vn(normal coordinates) and f(faces of a mesh), other keys are ignored)
  • .PLY (all encoding types(ascii and byte) are supported with limitation to only float type for data)
vertices, normals = cv2.loadPointCloud("teapot.obj")

Function cv::loadPointCloud returns vector of points of float (cv::Point3f) and vector of their normals(if specified in source file). To visualize it you can use functions from viz3d module and it is needed to reinterpret data into another format

vertices = np.squeeze(vertices, axis=1)
color = [1.0, 1.0, 0.0]
colors = np.tile(color, (vertices.shape[0], 1))
obj_pts = np.concatenate((vertices, colors), axis=1).astype(np.float32)
cv2.viz3d.showPoints("Window", "Points", obj_pts)
cv2.waitKey(0)

In presented code sample we add a colour attribute to every point Result will be:

For additional info grid can be added

vertices, normals = cv2.loadPointCloud("teapot.obj")

Other possible way to draw 3d objects can be a mesh. For that we use special functions to load mesh data and display it. Here for now only .OBJ files are supported and they should be triangulated before processing (triangulation - process of breaking faces into triangles).

vertices, indices = cv2.loadMesh("../data/teapot.obj")
vertices = np.squeeze(vertices, axis=1)
cv2.viz3d.showMesh("window", "mesh", vertices, indices)