Resistive network for detecting the centroid of nonlinear coordinates

For realizing visual feedback control, vision chips are promising. However early-vision-like processing is not sufficient because its output is pattern information which cannot be directly used in actuator control. We propose a two-dimensional resistive network for extracting the zeroth- and first-order moments of two-dimensional coordinates in parallel. The network can detect the moments of nonlinear coordinates, so that it can be used to directly detect an object position not only in an imager coordinate system but also in a working coordinate system, or used for foveated vision. Through its use, a high-speed smart sensor for visual feedback is realized.