Neuronal Processing of Optic Flow (eBook)

Markus Lappe (Herausgeber)

eBook Download: PDF

1999 | 1. Auflage
321 Seiten
Elsevier Science (Verlag)
978-0-08-085778-7 (ISBN)

Infos

When we walk, drive a car, or fly an airplane, visual motion is used to control and guide our movement. Optic flow describes the characteristic pattern of visual motion that arises in these situations. This book is the first to take an in-depth look at the neuronal processing strategies that underlie the brain's ability to analyze and use optic flow for the control of self-motion. It does so in a variety of species which use optic flow in different behavioral contexts. The spectrum ranges from flying insects to birds, higher mammals and man. The contributions cover physiological and behavioral studies as well as computational models. Neuronal Processing of Optic Flow provides an authoritative and comprehensive overview of the current state of research on this topic written by a group of authors who have made essential contributions to shaping this field of research over the last ten years. - Provides the first detailed overview of the analysis of complex visual motion patterns in the brain - Includes physiological, behavioral, and computational aspects of optic flow processing - Highlights similarities and differences between different animal species and behavioral tasks - Covers human patients with visual motion deficits - Enhances the reader's understanding with many illustrations

Inhalt

Front Cover 1
Neuronal Processing of Optic Flow 4
Copyright Page 5
CONTENTS 6
Contributors 10
Foreword 12
Preface 16
PART I: PERCEPTION 20
Chapter 1. Human Ego-Motion Perception 22
I. Introduction 22
II. Retinal Flow and Optic Flow 23
III. Basic Properties of Heading Perception 25
IV. The Rotation Problem 26
V. Special Visual Strategies to Solve the Rotation Problem 30
VI. Circular Heading and Curved Motion Path Percept 32
VII. Heading Perception and the Pattern of Flow 35
VIII. Temporal Properties of Heading Perception 37
IX. Heading Perception and Moving Objects 39
X. The Reciprocal Relation between Optic Flow and Ego-Motion 40
References 41
PART II: EYE MOVEMENTS 46
Chapter 2. Optic Flow and Eye Movements 48
I. Introduction 48
II. Gaze during Self-Motion 49
III. Ocular Reflexes during Self-Motion 51
IV. Optic Flow Induced Eye Movements 54
V. Implications of Eye Movements for Optic Flow Processing 61
VI. Conclusion 64
References 65
Chapter 3. The Role of MST Neurons during Ocular Tracking in 3D Space 68
I. Neuronal Activity in MST during Short-Latency Ocular Following 70
II. Neuronal Activity in MST during Short-Latency Vergence 76
III. Role of MST Neurons during Ocular Tracking in 3D Space 80
IV. Tracking Objects Moving in 3D Space 80
References 81
PART III: ANIMAL BEHAVIOR AND PHYSIOLOGY 84
Chapter 4. Visual Navigation in Flying Insects 86
I. Introduction 86
II. Peering Insects 87
III. Flying Insects 88
IV. Concluding Remarks 107
References 108
Chapter 5. Neuronal Matched Filters for Optic Flow Processing in Flying Insects 112
I. Introduction 112
II. Visually Guided Behavior and Optic Flow Processing in Flying Insects 113
III. How to Gain Self-Motion Information from Optic Flow 116
IV. The Fly Visual System 118
V. Mapping the Local Response Properties of Tangential Neurons 121
VI. Response Fields and Matched Filters for Optic Flow Processing 127
VII. Conclusion 130
References 134
Chapter 6. A Common Frame of Reference for the Analysis of Optic Flow and Vestibular Information 140
I. Object Motion versus Self-Motion 140
II. The Accessory Optic System 141
III. Conclusion 155
References 156
Chapter 7. Optic Flow and the Visual Guidance of Locomotion in the Cat 160
I. Introduction 160
II. Uses of Vision during Locomotion 161
III. Gaze during Visually Guided Locomotion 166
IV. Neural Mechanisms for Analyzing Optic Flow Information 169
V. Conclusion 185
References 186
PART IV: CORTICAL MECHANISMS 190
Chapter 8. Stages of Self-Motion Processing in Primate Posterior Parietal Cortex 192
I. Motion-Sensitive Areas in the Macaque Visual Cortical System 193
II. Cortical Vestibular Areas 210
III. Human Brain Areas Involved in the Processing of Self-Motion Information 211
IV. Conclusion 212
References 212
Chapter 9. Optic Flow Analysis for Self-Movement Perception 218
I. Introduction 218
II. MST Sensitivity to Heading Direction 219
III. MST Sensitivity to the Structure of the Environment 223
IV. MST Responses to Real Translational Self-Movement 226
V. Interactions between Optic Flow and Translational Self-Movement 229
VI. MST's Role in Self-Movement Perception 232
VII. A Distributed Network for Self-Movement Perception 233
References 235
Chapter 10. Neural Mechanisms for Self-Motion Perception in Area MST 238
I. Area MST–Optic Flow Selectivity 239
II. Area MST–Shifting Receptive Fields 243
III. Conclusion 250
References 250
Chapter 11. Computational Mechanisms for Optic Flow Analysis in Primate Cortex 254
I. Introduction 254
II. Foundations and Goals of Modeling 255
III. Models of Optic Flow Processing in Primates 257
IV. Comparisons with Physiology: Optic Flow Representation in Area MT 261
V. Comparisons with Physiology: Optic Flow Selectivity in Area MST 264
VI. Receptive Fields of Optic Flow Processing Neurons 273
VII. The Population Heading Map 275
VIII. Conclusion 283
References 284
Chapter 12. Human Cortical Areas Underlying the Perception of Optic Flow: Brain Imaging Studies 288
I. Introduction 288
II. New Techniques in Brain Imaging 293
III. Summary 308
References 309
Chapter 13. What Neurological Patients Tell Us about the Use of Optic Flow 314
I. Introduction 314
II. Functional Architecture of Motion for Navigation 314
III. Why Study Motion-Impaired Neurological Patients? 316
IV. The Radial Flow Field 318
V. Impairment of Locomotion and Recovery of Locomotor Function 321
VI. Heading Perception in the Presence of Objects 322
VII. Conclusion 329
References 330
Index 336

Textauszug

Human Ego-Motion Perception

A.V. van den Berg    Helmholtz School for Autonomous Systems Research, Department of Physiology, Faculty of Medicine, Erasmus University, Rotterdam, the Netherlands

I Introduction

A seemingly simple task like walking an empty corridor without hitting the walls becomes very difficult when asked to do so blindfolded. Toddlers who have just learned to walk tip over when the walls of a movable room are set into motion (Stoffregen et al., 1987). Walking on a treadmill that is towed around at speeds different than the treadmill’s speed result in changes of the felt walking speed (Rieser et al., 1995). These examples illustrate that the interplay between vision, kinaesthetic, and vestibular information is of major importance to the control of locomotion.

In order to serve locomotion, the visual system needs to represent ego-motion in a format that is useful to act in the environment. Thus, one needs to specify what sort of visual information is relevant to locomotion and if—and how—this visual information is acquired. Because locomotion is a broad description of many different tasks that require different elements of visual information (e.g., walking toward a target, making a turn, and avoiding obstacles), the required visual information is to some extent task-specific. For example, to prevent bumping into an obstacle, it is useful to perceive whether it is on one’s future path and how much time is left for corrective action. The distance to the object is not relevant except in proportion to the speed of forward motion. Consequently, much attention has been given in the psychophysical literature to the visual perception of heading and judgments of the time to contact. In this review, I will concentrate on the first of these tasks: the perception of heading.

Gibson (1966, 1986) recognized that the visual motion field contains useful information for such tasks. He observed that the pattern of direction lines that connects a vantage point with objects in the environment expands when the vantage point moves forward. Only that direction line that coincides with the direction of forward motion remains stationary. Thus, the moving vantage point receives an expanding motion pattern that radiates outward from the direction of heading. This pattern of motion is called the optic flow, and its center is called the focus of outflow. In Gibson’s view, the focus of outflow labels the object or the location in the environment to which one is heading. There is no need for a specification of a reference frame for the measured flow. The array of visual objects serves as the frame with respect to which the heading direction is visually specified. These ideas of Gibson have served as a useful starting point for the analysis of visual perception of heading. One can find an excellent review of older literature in Warren (1995).

II Retinal Flow and Optic Flow

Even when the observer is moving on a linear track, the flow on the retina will rarely be a purely expanding motion pattern. This holds because the retina is placed on top of a series of mobile supports (e.g., the hips, the torso, the head and the eye), which can all rotate relative to one another. It is useful therefore, to make a clear distinction between retinal and optical flow, the former depending on the translational and the rotational movements of the eye, whereas the latter only involves the translatory component of the eye. Both types of flow fields are typically represented by a collection of angular motion vectors, each attributed to a particular visual direction line (Fig. 1). This representation of the flow field is appropriate for heading analysis (Warren et al., 1991a), but derivatives of the flow field may be more appropriate for other tasks like shape from flow (Koenderink, 1986).

Fig. 1 The retinal motion pattern depends on the pattern of motion on the screen and the eye’s rotation. If the motion pattern on the screen simulates the eye’s approach of a wall (upper panel), the effect of the eye rotation will be to shift the center of expansion on the retina relative to the center on the screen. One of the moving dots on the screen will be stable on the retina, whereas the dots that correspond to the focus on the screen will be moving relative to the retina. The shift on the retina will be in the same direction as the eye’s rotation (left and right panels). Its magnitude depends on the simulated speed of approach, the eye’s rotation and the simulated distance to the wall. If the simulated scene is not a wall, there may be no clear focus on the retina, yet there may be an apparent focus that is consistent with a “best fit” of an expanding flow field to the actual retinal flow.

The eye’s translation causes angular motion away from the direction of heading with a magnitude that is inversely proportional to the distance. The eye’s rotation generates flow that consists of parallel motion across the retina with a magnitude that is independent of the distance. Its direction and magnitude merely depend on the orientation of the axis of rotation and the rotational velocity. More importantly, it does not even depend on the location of the rotational axis relative to the eye. This gives rise to an ambiguity in the relation between the instantaneous flow field and the eye’s motion through the environment. Moreover, the rotations usually change over time in direction and magnitude as does forward motion, leading to nonstationary flow. Yet, current research has mostly dealt with stationary flow patterns (but see Cutting et al., 1992). For the moment, we ignore these difficulties and discuss various studies that have dealt with heading perception from pure expanding retinal motion.

III Basic Properties of Heading Perception

Studies of ego-motion perception have greatly profited from the advent of affordable fast graphics workstations that can simulate 3D scenes in real time. Typically, one simulates the retinal flow for an eye that moves through scenes without recognizable features (randomly located dots). Such patterns may evoke vivid perception of self-movement, called linear vection. Vection latency and strength depend on the display size, type of flow, direction of simulated motion (Telford and Frost, 1993) and the richness of motion in depth cues (Palmisano, 1996). Linear vection takes several seconds to build up, but the percept of ego-motion direction or heading occurs well within a second (Crowell et al., 1990; Warren and Kurtz, 1992; Crowell and Banks, 1993; Stone and Perrone, 1997), even when the sense of self movement is still relatively weak.

Simple simulations in heading studies involve motion of an eye on a linear track. This turns out to be a relatively simple task if the eye fixates some stationary target on the screen, resulting in pure retinal expansion. Heading can then be discriminated from a reference target in the scene with a just noticeable difference (jnd) angle of 1–2° (Warren et al., 1988), which is thought to be sufficient for avoidance of obstacles during normal locomotion (Cutting et al., 1992; Cutting, 1986). This performance level is little affected by changes in the layout of the simulated scene (Warren et al., 1998; te Pas, 1996), the presentation time (down to 300 ms: Crowell et al., 1990; down to 228 ms: te Pas et al., 1998) or density of the simulated environment (down to 3 visible dots: Warren et al., 1988). Also, the retinal locus of the simulated heading does not affect discrimination performance very much although there is an accuracy gain of the central region over the periphery (Warren and Kurtz, 1992; Crowell and Banks, 1993; te Pas et al., 1998). Azimuthal and elevational components of heading may have different retinal loci of optimal discriminability. Azimuthal precision is slightly larger in the lower hemi-retina than in the upper half (D’Avossa and Kersten, 1996). In contrast to these rather mild effects of retinal location, there is a clear penalty paid when the focus is off-screen. If the flow within a small aperture is nearly parallel (because the focus is very eccentric), finding the focus of the flow vectors is strongly affected by noise of the visual processing (Koenderink and van Doorn, 1987). Indeed, the jnd between two heading directions increases by nearly two orders of magnitude (up to about 30°) when the focus is moved out from the center of a 10° diameter display to 60° eccentricity (Crowell and Banks, 1993).

Thus, consistent with Gibson’s hypothesis, the pattern of expanding flow vectors provides the information for heading direction, and the well-known retinal inhomogeneity has a relatively minor effect on the performance.

IV The Rotation Problem

Of course, the eye often rotates relative to the environment as we habitually turn our eyes and/or head to pursue targets in our environment or because we are moving on a curved trajectory. This adds a rotational component to the expansion flow, which destroys the focus at the direction of heading. For special layouts of the environment, like an extended wall, a new singular point appears in the direction of eye rotation (Fig. 1). Responding to this...

Produkt-Details

• Erscheint lt. Verlag: 6.12.1999
• Mitarbeit: Herausgeber (Serie): Ronald J. Bradley, Robert Adron Harris, Peter Jenner
• Sprache: englisch
• Themenwelt: Medizin / Pharmazie ► Medizinische Fachgebiete ► Augenheilkunde
• Themenwelt: Medizin / Pharmazie ► Medizinische Fachgebiete ► Neurologie
• Themenwelt: Studium ► 1. Studienabschnitt (Vorklinik) ► Physiologie
• Themenwelt: Naturwissenschaften ► Biologie ► Humanbiologie
• Themenwelt: Naturwissenschaften ► Biologie ► Zoologie
• ISBN-10: 0-08-085778-7 / 0080857787
• ISBN-13: 978-0-08-085778-7 / 9780080857787