Path perception from optic flow

Abstract

Perceiving the path we are travelling on is important for successful navigation. Relative motion between the world and the observer generates optical flow on the retinae (retinal flow). Gibson (1950) pointed out that when travelling on a straight path with no eye, head, or body rotation, retinal flow is radial and the stationary point indicates the instantaneous direction of travelling, or heading, of the observer. The straight path can then be recovered as it coincides with heading.

Nevertheless, it is rarely the case that people travel with no rotation. Instead, they normally look at different points of interest when they are navigating. The result of changing one's gaze or rotating one's head is the addition of a rotational component, which is a laminar flow, to the flow field. The rotational component shifts the stationary point from heading and makes heading perception difficult. Extensive research has been conducted on how the human visual system removes the rotational component of the retinal flow and how extra-retinal information, such as efferent copies of eye muscle commands, may contribute to this process.

The paths on which people travel are not always straight, but often curved. When a path is curved, it no longer coincides with heading. In this case, heading is the tangent of the path. Researchers have proposed theories to explain how curved paths are perceived. Each of them requires different visual information and gaze conditions (e.g., fixating on a target or gazing along the heading direction). They can be categorized by whether or not path perception depends on heading perception.

The goal of this thesis is to systematically examine different theories of path perception and determine how humans perceive curved paths. Study 1 examined different path perception theories by comparing human path perception performance in various gaze conditions and with the availability of various optic flow information. Study 2 investigated whether path perception depends on heading perception. Study 3 examined the contribution of reference objects to path perception. Study 4 investigated how extra-retinal informationcontribute to path perception.

The experiments that I present here show that (a) when there is no extra-retinal information, path perception is accurate only when one's gaze is along heading such that the rotation in the flow field is equal to path rotation; (b) when one's gaze is not along heading such that the rotation in the flow field is not equal to path rotation, path perception is inaccurate. Adding more visual information, such as acceleration, dense flow field, and/ or reference objects does not improve the accuracy; (c) eye movement signals support accurate path perception only in the natural case of self-motion in which body orientation is aligned with heading such that eye movement signals help to stabilize heading in the body-centric coordinate system.