Occlusion, Transparency, and Depth Effects in Visual Search for Laboratory and Familiar Real-World Objects

Rachel Nguyen

Advisor: Matthew S. Peterson, PhD, Department of Psychology

Committee Members: William S. Helton, Martin Wiener

Online Location, Online
April 19, 2024, 01:00 PM to 03:00 PM

Abstract:

When visual search occurs in unnatural scenes (e.g. TSA baggage screening), we do not have contextual cues from top-down processing to aid search (e.g. initial searches for a laptop in an office may occur on top of a desk or in a bookbag), and in crowded scenes, the chance of occlusions occurring between targets and distractors increases, and occlusion affects our perception of objects. When laboratory objects are not fully visible, search is inefficient if there is a gap of visible information within an object, and search is even more inefficient when there is a distractor object on top of a target. To explore whether these visual search results are similar between laboratory and real-world objects, Study 1 (Experiment 1) examined if the type of missing visual information affected feature search of familiar real-world objects (Hershey Bar and a Remote Controller). In addition, Study 1 (Experiment 2) also tested whether there were differences between feature and identity searches (the latter is more typical in real-world scenarios). Bayesian analyses between occluder and stimuli type resulted in stronger evidence that search for complex objects is more effortful, and missing visual information was disruptive. White occluders affected feature searches more than identity searches. To determine if target depth and distractor object opacity levels drive search (when occlusion is present), Study 2 (Experiment 1) addressed if search for targets “above” a distractor was easier than search for targets “below,” and this study also addresses how search is affected by different distractor opacity levels (0%, 25%, 50%, 75%, 100%). To test whether an irrelevant feature (color) affects orientation search, Study 2 (Experiment 2) participants searched for a vertical bar (amongst horizontal bars) that was light blue or light yellow (colors sensitive to detection). Bayesian analyses show that the best model predictor for search of transparent stimuli includes depth and diagonal opacity--not including color. To test whether search for real-world targets was affected by depth and occluder opacity compared to a simple object, participants searched for Hershey bars or Remote Controllers (Study 3, Experiment 1) and grayscale-Hershey bars or grayscale Remote Controllers (Study 3, Experiment 2) with the same design and conditions as Study 2. Bayesian analyses show strong-to-decisive evidence for easier searches when objects are “above” a distractor, for differences between simple and complex real-world objects occuring in search for targets below a distractor object, and for differences between complex, real-world objects occur when targets are below and distractors are at 50% (and higher percentages) of opacity. Finally, Study 4 addresses how occlusion (0%, 25%, 50%, 100%) affects search for real-world objects when familiar real-world object distractors can partially occlude the target. In addition, stimuli type was manipulated between subjects (Normally colored, Grayscale, Transparent-Colored, Transparent-Grayscale). Since Study 1 and 3 results show evidence that occlusion decreases search efficiency, this study also addresses whether increasing levels of occlusion affect search efficiency. For naturally colored objects, transparency made search harder compared to opaque colored objects. For grayscale objects, there were no search efficiency differences between opaque and transparent objects. Thirdly, as a follow-up to study 3 on depth effects for object identity search, analysis on relative target-to-distractor layer effects was also conducted, and there were significant differences in search for targets across two layers. Overall, when searches occur in congested scenes where occlusion is prevalent, depth and occlusion type affects feature and identity search for laboratory and complex, real-world objects. In addition, color seems to aid search for opaque, real-world objects.