An Analysis of the Motion Signal Distributions Emerging from Locomotion through a Natural Environment

摘要: Some 50 years have passed since Gibson drew attention to the characteristic field of velocity vectors generated on retina when an observer is moving through three-dimensional world. Many theoretical, psychophysical, and physiological studies demonstrated use such optic flow-fields for a number navigational tasks under laboratory conditions, but little known about actual flowfield structure natural operating conditions. To study motion information available visual system in real world, we moved panoramic imaging device outdoors accurately defined paths simulated biologically inspired detector network analyse distribution signals. We found that signals are sparsely distributed space local directions can be ambiguous noisy. Spatial or temporal integration would required retrieve reliable vectors. Nevertheless, surprisingly simple algorithm rather direction heading from sparse noisy signal maps without need pooling. Our approach thus may help assess role specific environmental computational constraints flow processing.