| Budget Amount *help |
¥12,000,000 (Direct Cost: ¥12,000,000)
Fiscal Year 1985: ¥5,200,000 (Direct Cost: ¥5,200,000)
Fiscal Year 1984: ¥6,800,000 (Direct Cost: ¥6,800,000)
|
|---|
| Research Abstract |
Although we human beings are not conscious in every instant, we employ so many kinds of visual attentions during task execution, to find objects, to monitor environment changes, to guide the motion of objects, and to verify environment situations before or after some actions. If we could provide a robot with fast and flexivle capability of such cisual attentions, robot-environment interactions would be improved very much. The vision programs for those interactions can be constructed to focus visual attentions onto local regions where key features are expected to appear. Generally, points of attention should be multiple and run in parallel. In order to realize such visual attentions, we designed and implemented a versatile multi window vision system for robots. A rectangular local region to be processed for one attention is referred to as a window. It is not fixed on the screen. Location, shape, and spatial data sampling rate for individual windows can be vontrolled independently and simultaneously.
In this reaearch, we first designed and developed a prototype of multi window vision system by using TTL LSIs. Then, we designed semicustom LSI for the electronic circuits of above mentioned window control logic and implemented it onto CMOS gate array of 2400 gates in 40 pin package. By using that LSI, we also developed the multi window vision system that has 56 windows. Experiments prove that the system provides fast and flexible capability for robot vision applications.
|
