Lighting Control
| Lighting in the KidsRoom was designed with two purposes in mind. First, the computer vision tracking and recognition algorithms require that the room be well-lit from above and that the lighting settings can be reliably set prior to each run. Second, the lighting is used in transition scenes to make the room seem more magical. The lighting is fully computer controlled. |
Vision Constraints
Before the KidsRoom is run, images of the room, emptied of people, are digitized. During the run, whenever the vision routines are active the lighting must be exactly the same as when the background images were recorded. The tracking and recognition algorithms require that the room be brightly lit from above and lit as evenly as possible. Special blinders were used to block light from falling on the screens, which look best when dark. When the lighting in the room changes during the transitions scenes (e.g. flashing colored lights), the vision algorithms are shut down until the lighting returns to the stable, bright settings. Eleven white theater lights were arranged to light the floor of the room as consistently as possible. Still, there is enough variation in lighting so that the tracking algorithm had to be somewhat intensity invariant, since brightness changes from one side of the room to the other.
MIDI Control
Lighting control works as follows. The control program sends a command to a process running on an SGI Indy workstation. That process translates the command into MIDI light signals and sends them to a light board. The light board has been programmed with several standard lighting sequences, such as raise-main-lights, lower-lights, blue-light-on, etc. One of these sequences is activated by the MIDI command and the lights respond.
Transitions
Three colored lights provide mood lighting during the transition scenes from the bedroom to the forest and from the MonsterLand back to the bedroom. Combined with "transitional" music, the flashing lights give a nice sense that something important is happening. During these transitions, however, the vision algorithms are deactivated.
