From Optics to ActionsFrom Optics to ActionsFrom Optics to ActionsJoe Lappin7 April 2004Joe LappinJoe Lappin7 April 20047 April 2004A flowing optic arrayGeneral theme: The dynamics of an observer’sactions — their timing, direction, and force — areeffectively controlled by the continuously changingoptical produced by the observer’s movements. • Why is this interesting? important? • What is the optical information? • What are the visual mechanisms?Reading:• Warren, W.H. (2004). Optic flow. In J.S. Werner & L.M. Chalupa (Eds.), The visualneurosciences (Ch. 84, pp. 1247-1259). Cambridge, MA: MIT Press.• Also relevant: Duffy, C.J. (2004). The cortical analysis of optic flow. (Ch. 85 inthe same book).Examples: • Walking, running, skiing, skating, etc.(Demand accuracies of ~ 1-3 deg.) • Driving cars, riding bikes, flying planes • Games with fast-moving balls(Regan et al., 1995, calculate that bat-ball intersections incricket & baseball must occur within a space-time window of about 10cm x 2.5 msec!) • Anticipating & controlling collisions, time-to-contact: τ = θ/dθ/dt. (David Lee’s research with long-jumpersand diving birds — indirect evidence of optical control of timing.Yilmaz & Warren (1995) provided more direct evidence inexperiments on braking in response to computer-generated displays.)Brief review of some principal ideas: • Gibson (1950) coined the term “optic flow” to refer to the changingstructure of optical images produced by observers’ movements through 3Denvironments.• visual importance of changing rather than static images• visual importance of the active observer• Locomotion (e.g., of pilots in planes) must be controlled by the changingstructure of optical patterns.• focus of outflow (of expansion, FOE) specifies heading direction • Gibson (1979): “ecological approach” — perception, environment, & actionas inseparable • David Lee (e.g., 1980, Proc. Royal Society) develops theory and evidenceto show that the rate of optical expansion, tau, τ = a/a*, provides visibleinformation about time-to-contact. • Differential geometry of optic flow (Koenderink & van Doorn, 1970s,1986): 1st-order differential components: div [divergence, expansion], curl[rotation in image plane], def [deformation, slant]Issue: Optic flowor retinal flow? Can the expansioncomponent and headingdirection be extractedfrom retinal imagesproduced by bothobserver translation andeye movements? Retinal coordinates orrelative motions & imagedeformations? RFs of some MSTdcells change with activeeye movements.MST neurons have very largeRFs, often > 40˚ diameter.Thus, they will integrate largeareas of optic flow.Differentiation (via center-surround antagonism) also occursin both area MT and MST.The images at the right showthe complex RF of a neuron inMSTd. (from Duffy, 2004)Thank you for your
View Full Document