David S. Young

Visual timing in hitting an accelerating ball.

To investigate the timing of actions relative to events in the environment, we observed subjects leaping to punch a falling ball. We analysed their knee and elbow angles as functions of time for three ball-drop heights, finding that the differences in the functions for the different heights could be explained on the basis that the subjects were gearing their actions to a particular optic variable. This variable specifies the time remaining before contact with an object if the closing velocity is constant; for the falling ball it gives an increasingly accurate estimate of the time-to-contact. Our visuo-moto control model incorporates a delay parameter, the value of which was estimated from the data. In addition, correlations indicated that the knee and elbow were generally quite tightly coupled. The relationship of this task to laboratory tracking tasks and to the timing of actions in everyday life is described.

Visual control of step length during running over irregular terrain.

Running over uneven ground requires visually regulating step length to secure proper footing. To examine how this is achieved, we studied subjects running on a treadmill on a series of irregularly spaced targets. The movements of their lower limbs and coccyx relative to the targets were monitored opto-electronically by a Selspot system. The results indicated that step length was adjusted to strike the targets primarily by varying the vertical component of impulse applied to the ground during the stance phase. In contrast, horizontal impulse was not varied significantly, and changing the reach forward of the foot on landing contributed little to variation in step length. Changing the vertical impulse simply altered the step time proportionately. Thus the data are consistent with a time-based model in which vertical impulse is modulated by the optic variable delta tau (Lee, 1976) specifying the time gap that has to be bridged by the runner between two targets.

Visual Timing of Interceptive Action

One of the most remarkable aspects of motor skill is the precision with which actions can be timed. In this chapter we will be concerned with the control of actions whose timing is dictated by how the organism is moving relative to the environment, as in locomotion, or by how an object is moving relative to the organism, as in catching or hitting something. We will refer to this as extrinsic timing. It requires predictive information about the relative motion of the organism and objects and surfaces in the environment, information that is often only available through vision. Time to contact (meaning, in general, time to nearest approach) is a particularly important predictor.

A roadside simulation of road crossing for children

A simple and safe method is proposed for giving children practical experience similar to crossing the road and for assessing their performance. The method comprises a ‘pretend road’ laid out on the pavement, which the child crosses as if crossing the adjacent road in the face of oncoming vehicles. A comparison of adult performances in crossing through gaps in traffic on pretend and real roads indicates that the pretend task adequately simulates real road crossing. Similar experiments on pretend roads with 5-10 year olds crossing through gaps in traffic indicated that children are able to understand the simulation task and perform sensibly on it. Moreover, there were children of all ages who consistently performed at close to an adult level indicating that the visual timing required in the skill is not beyond young children. The results suggest that children should be trained in crossing in the presence of traffic at an early age. The pretend task could prove a valuable aid to training.

How do somersaulters land on their feet

Body movements of trampolinists landing upright from forward somersaults, with eyes open and closed, were analyzed to test a theory of how braking and timing of actions are conjointly controlled. In regulating landing, rotation has to be slowed by extending the body so that it reaches the upright just as the feet hit the trampoline. Extending a theory of visual control of linear braking, it was hypothesized that upright landing is achieved by perceptually regulating body extension so as to keep the ratio angle of body to upright: rate of change of angle (the tau function of the angle) proportional to time-to-landing. The data support the hypothesis, indicate that vision improves precision of control, and argue further for the value of the tau function in the perceptual regulation of action.

Straight lines and circles in the log- polar image

Foveal or spatially-variant image representations are important components of active vision systems. Log-polar sampling is a particularly powerful example as a result of the simplicity with which expansion and rotation can be handled. These properties are exploited here for the detection of general straight lines, line segments, and circles through the foveation point. An efficient and practical method based on convolution is described, and investigated in the context of a simple foveation strategy.

First order optic flow from log-polar sampled images

The first-order spatial derivatives of optic flow — dilation, shear and rotation — provide powerful information about motion and surface layout. The log-polar sampled image (LSI) is of increasing interest for active vision, and is particularly well-suited to the measurement of local first-order flow. We explain why this is, propose a simple least-squares method for measuring first-order flow in an LSI sequence, and demonstrate that the method works well when applied to real images.

Dynamic fixation of a moving surface using log polar sampling

We describe the development and testing of a first-order motion estimation algorithm which maintains accurate fixation of features on surfaces undergoing three-dimensional motion, and determines the local affine motion parallax. The accuracy of the first-order flow estimation is much improved by the use of log-polar sampling. We investigate the contribution of fixation to this accuracy using synthetic flow, and demonstrate the performance on affine tracking in real image sequences.

Building a Model of a Road Junction Using Moving Vehicle Information

We describe a program to construct a model of a road junction using data from a single camera. The model specifies the ground plane orientation in camera coordinates and the positions of traffic lanes, and is obtained entirely from observations of vehicle movements, with no static image analysis. At present. the model is restricted to representing straight lane segments. We describe our methods for segmentation. object tracking, ground plane estimation and lane identification. Throughout. we emphasise techniques which are computationally cheap and can be used with fairly low resolution data and a low frame rate. We nevertheless obtain a reliable model by using the statistics of large numbers of vehicle movements.

Image Retrieval Methods for a Database of Funeral Monuments

This paper investigates the use of Gabor features in matching 2- and 3-D objects in photographic images, using a database of images of funeral monuments from English churches. The technique, which can be applied to large databases, allows an arbitrary patch of a reference image to be matched to patches of each database image at a range of scales and positions. We investigate the use of nonlinear preprocessing to reduce the influence of lighting and surface reflectance on the match results.