Only released in EOL distros:
3d_interaction: blob3d | external_camera_localizer | projected_lightswitch_interface | projector_calibration | projector_interface | rcommander_interface | world_intersect
Package Summary
Documented
Provides a node for calibrating a projector to a camera with a known TF frame.
- Author: Daniel Lazewatsky
- License: BSD
- Source: git https://github.com/OSUrobotics/ros-3d-interaction.git (branch: groovy-devel)
Nodes
grid
This node takes in a camera image and computes the homography between a detect grid and the idealized grid that it is projecting.Subscribed Topics
image (sensor_msgs/Image)- A camera image containing the projected grid to be detected.
Published Topics
grid (sensor_msgs/Image)- The grid which is being projected
- The homography, mapping points in the camera image to projector pixels.
Parameters
~grid_size (string)- Defaults to 5x7
Using the Calibration
Given a homography, H (provided by projector_calibration/grid), and camera calibration K (from a sensor_msgs/CameraInfo), a 3D point, p in the world can be projected onto as follows:
Transform p into the same frame as the camera used to find H (call the new point pimg).
Project pimg onto the image plane (call the new point pproj).
Transform p2d to projector pixels using H
In python this looks something like
1 import image_geometry, tf, cv2
2
3 # get camera_info_msg and H from somewhere
4 # create a tf listener
5
6 model = image_geometry.PinholeCameraModel(camera_info_msg)
7 p_img = tf_listener.transformPoint(model.tf_frame, p)
8 p_2d = model.project3dToPixel((p_img.x, p_img.y, p_img.z))
9 p_proj = cv2.perspectiveTransform(H, [p_2d])
