Abstract

Motion is an important component of vision for flying insects. Optomotor reactions serve to minimize involuntary deviations of trajectories and stabilize the flight course. Insect optomotor responses are based on directional movement detection. The Reichardt correlation detector, essentially based on linear time-invariant filtering, cannot account for the saturation and adaptation properties exhibited by insect responses to motion. In this paper a directionally selective movement detecting (DSMD) neural network architecture, inspired by invertebrate vision, is presented. It is structurally and, to second order, functionally equivalent to a Reichardt detector, but it operates on the nonlinear mechanism of shunting lateral inhibition. Responses, under various constraints, of this DSMD neural network to moving patterns such as edges, bars, and sinusoidal gratings are investigated and compared to those of insect DSMD interneurons.

© 1992 Optical Society of America