Depth camera sensor
This sensor can be used to simulate laser range sensors. It uses the z-buffer of
a Blender camera to generate a depth image and then process it to produce a list
of 3D points, given with respect to the position of the camera.
The cameras make use of Blender’s bge.texture module, which requires
a graphic card capable of GLSL shading.
Also, the 3D view window in Blender must be set to draw Textured objects.
This sensor requires the generation of z-buffer images instead of the usual
RGBA. For this reason, exporting the images will result in strange (and quite
Psychodelic) colours.
The settings for the near and far planes of the camera are important
for this sensor. They will determine the resolution of the depth data
recovered. For best results, it is recommended to keep the near plane as
far as possible, and the far plane as close as possible. Reducing the range
between both will result in higher resolution, specially towards the far
plane. For more information:
http://www.codermind.com/articles/Depth-buffer-tutorial.html
Note
The streaming of data from this sensor can be toggled off and on by
pressing the SPACE key during the simulation. This will affect all the
video cameras on the scene.
Toggling off the cameras can help make the simulation run faster,
specially when there are several cameras. However, the lack of
data on the stream may cause problems to some middlewares.
Files
- Blender: $MORSE_ROOT/data/sensors/depth_camera.blend
- Python: $MORSE_ROOT/src/morse/sensors/depth_camera.py
Local data
- 3D_points: A list of lists. Each element of the list has the format
[x, y, z], with the coordinates given with respect to the origin of
the camera.
- intrinsic_matrix: The intrinsic calibration matrix, stored as a 3x3
row major Matrix
Configurable parameters
The Empty object corresponding to this sensor has the following parameters:
- capturing: (Boolean) flag that determines whether the camera should
generate an image. It can be toggled on or off by pressing the Space.
The default value should be True.
- cam_width: (int) generated image width in pixels. The default value is
128.
- cam_height: (int) generated image height in pixels. The default value is
128.
- cam_focal: (double) camera focal length as defined in Blender (note: in
Blender this parameter unit is “millimeters”. This is actually misleading, as
there is no dimension associated to Blender units.) The default value is
35.0.
- cam_near: (double) Adjust the near_clipping parameter of the camera
(Anything closer to the camera than near_clipping is not displayed). The
default value is 1.0 (meter).
- cam_far: (double) Adjust the far_clipping parameter of the camera
(Anything further from the camera than far_clipping is not displayed). The
default value is 20.0 (meter).
- Vertical_Flip: (Boolean) flag that determines whether the image should be
flipped. Default value is False.
- Depth: (Boolean) flag that determines whether the sensor generates
z-buffer images or plain images. It should be set to True for this sensor
to work properly. In this case, the images generated will not be really
Visible, since the data will be stored as floats, instead
of the usual RGBA strings. Default value is True.
Camera calibration matrix
The camera configuration parameters implicitly define a geometric camera in
blender units. Knowing that the cam_focal attribute is a value that
represents the distance in Blender unit at which the largest image dimension is
32.0 Blender units, the camera intrinsic calibration matrix is defined as
alpha_u |
0 |
u_0 |
0 |
0 |
alpha_v |
v_0 |
0 |
0 |
0 |
1 |
0 |
where:
- alpha_u == alpha_v = cam_width . cam_focal / 32 (we suppose
here that cam_width > cam_height. If not, then use cam_height in
the formula)
- u_0 = cam_height / 2
- v_0 = cam_width / 2
Services
- capture: (asynchronous service) the method expects an integer n in
input and answer back when the simulated camera has token n shots.
Applicable modifiers
No camera modifiers available at the moment