Game Development Reference
In-Depth Information
Figure 2.4. Blind spots in the depth image caused by the offset of the IR transmitter
and receiver.
It is worth noting the location of the IR transmitter and receiver with respect
to one another, as well as the color camera. The IR transmitter is located at
the far side of the Kinect, with the color and IR cameras located in the center
of the device. The relative locations of these components on the Kinect have
a significant effect on their respective operations. For example, since the IR
transmitter is offset from the IR camera, the portions of the scene that can
be “viewed” by each of them are slightly different. This effect is depicted in
Figure 2.4, where it can be seen that there are portions of the scene where no
depth information is available.
In a real-time rendering context, you could imagine a very similar configu-
ration with a camera and a spotlight light source that are oriented in a similar
fashion as the IR transmitter and receiver are. The blind spot corresponds to
the “shadow” produced by the spot light, and the camera is still able to “see” a
portion of the shadow. This same effect applies to the relationship between the
depth camera and the color camera as well. There will be portions of the scene
that are visible in the color image that aren't visible in the depth image and vice
versa. In addition, this also means that a scene point within one image may or
may not be at the same pixel location within the other image. Since the two
cameras are close together these discrepancies are usually minimal, but they still
exist and must be taken into consideration. We will consider how to handle these
effects in Section, 2.3 , “Mathematics of the Kinect.”
Similar to the color-image stream, the depth data is made available to the
application in a variety of resolutions and data formats. The depth data itself
provides a 13-bit value representing the camera space Z-coordinate of the object
being sensed at each pixel. This value provides a millimeter precision value,
with a valid data range of either 800 mm to 4,000 mm or 400 mm to 3,000 mm
depending on the operational mode selected. In addition, the remaining 3 bits
per pixel can be used to provide a player ID produced by the Kinect skeletal
system. The available resolutions are 640
60. Once
again, please check the SDK documentation for full details about the possible
combinations of these options.
480, 320
240, or 80
Search Nedrilad ::

Custom Search