Teaching the visual system to segment and interpret images of overlapping transparent objects

Behavioural and eye-movement data gathered from five experiments that examined visual search behaviour/performance in stereoscopic displays. Experiments 1-4 involved examining search of stereoscopic displays under varying levels of overlap (0%, 45%, 90%), and varying task conditions (participants were asked to search for either one or two targets). Each of the first four experiments involved search for a different stimulus type (covering opaque polygons, transparent polygons, household objects and X-ray images respectively). Experiment 5 involved training participants to search through X-ray images for a series of sessions either in the presence of stereoscopic depth, or in flat displays, following by a transfer session to flat displays. Experiments 1-4 involved eye-tracking and Experiment 5 involved eye-tracking every four sessions, as well as in the transfer session.<p>A considerable amount of research has examined the problems associated with searching for multiple target objects, with a particular focus on X-ray baggage screening, where security personnel must search for a number of different objects simultaneously.

The goal of this project is to extend the previous research that has been conducted by examining the complexities associated with X-ray images, such as transparency and overlap. Security search for threat items (weapons) within X-ray images of baggage presents a challenge for the visual system because X-ray images do not conform to the conventional rules of solidity, interposition and opacity. In X-ray images, transparency causes colour changes at points of overlap. The colours representing each object interact at areas of overlap in a way that is largely unfamiliar to the visual system.

In this project, a series of experiments will be conducted using eye-tracking methodology to understand and explore the complexities involved in searching displays of this type. Furthermore, the displays will be presented in both 2D and 3D, to determine whether any difficulties associated with searching transparent overlapping images can be overcome when the objects are separated into different layers of depth.</p>