This thesis addresses the generation of behaviourally useful, robust representations of the sensory world in the context of machine vision and behaviour. The goals of the work presented in this thesis are to investigate strategies for representing the visual world in a way which is behaviourally useful, to investigate the use of a neurally inspired early perceptual organisation system upon high-level processing in an object recognition system and to investigate the use of a perceptual organisation system on driving an object-based selection process. To address these problems, a biologically inspired framework for machine attention has been developed at a high level of neural abstraction, which has been heavily inspired by the psychological and physiological literature. The framework is described in this thesis, and three system implementations, which investigate the above issues, are described and analysed in detail. The primate brain has access to a coherent representation of the external world, which appears as objects at different spatial locations. It is through these representations that appropriate behavioural responses may be generated. For example, we do not become confused by cluttered scenes or by occluded objects. The representation of the visual scene is generated in a hierarchical computing structure in the primate brain: while shape and position information are able to drive attentional selection rapidly, high-level processes such as object recognition must be performed serially, passing through an attentional bottleneck. Through the process of attentional selection, the primate visual system identifies behaviourally relevant regions of the visual scene, which allows it to prioritise serial attentional shifts towards certain locations. In primates, the process of attentional selection is complex, operating upon surface representations which are robust to occlusion. Attention itself suppresses neural activity related to distractor objects, while sustaining activity relating to the target, allowing the target object to have a clear neural representation upon which the recognition process can operate. This thesis concludes that dynamic representations that are both early and robust against occlusion have the potential to be highly useful in machine vision and behaviour applications.

In the study of covert visual attention, two event-related potential (ERP) components have been identified by earlier research. The N2 posterior contralateral (N2pc) component has been suggested to index the enhancement of attention to a specific lateralized target item. The distractor positivity (PD) component has been suggested to index the suppression of distractors appearing in the same search array. Earlier studies have reported different latencies for the PD component depending on the task and experiment. Furthermore, the N2pc and the PD component are not always elicited in the same experiment. Relative target-to-difficult-distractor placement have been shown to affect the mean amplitude of the N2pc. Less is known about how different relative placements affect the PD component. The aim of the present study was to try and elicit both an N2pc and a PD component in the same visual search paradigm. The PD was recorded later time-window which previous studies have suggested to indicate the ending of attention to a previously attended target. Three relative placements were analysed, horizontally opposite, vertically opposite and diagonally opposite. When combining all three relative placements an N2pc component was elicited contralateral to the target. No PD component was found when combining all relative placements. A larger mean amplitude N2pc was measured for the vertically opposite condition. The results are not in line with previous research, that have found the N2pc to be smaller in conditions where both target and distractor are on the same side of the visual field. However, when comparing upper and lower visual field targets the N2pc was found to be larger for lower visual field stimuli, which is in line with previous research. A larger mean amplitude for the PD was found in the diagonally opposite condition. Earlier research has suggested that when difficult distractor and target are located on separate sides of the visual field, this leads to successful inhibition, indexed by the PD component. In contrast to earlier research a larger PD component was not found for upper visual field stimuli. The present study differs from previous studies in the way the target and difficult distractor were placed and analysed. By separating what has previously been called “opposite side” condition into two separate conditions diagonally opposite and horizontally opposite the results from the present study seem to suggest that these two conditions are not synonymous. However, the results should be regarded with caution due to the small sample size. Furthermore, the horizontally opposite side condition also differs from previous studies with regards to relative target and distractor distances, which could have had an effect on the results.

In this thesis, we study object ranking in mobile 3D visual search. The conventional methods of object ranking achieve ranking results based on the appearance of objects in images captured by mobile devices while ignoring the underlying 3D geometric information. Thus, we propose to use the method of mobile 3D visual search to improve the ranking by using the underlying 3D geometry of the objects. We develop an algorithm of fast 3D geometric verication to re-rank the objects at low computational complexity. In that scene, the geometry of the objects such as round corners, sharp edges, or planar surfaces as well as the appearance of objects will be considered for 3D object ranking. On the other hand, we also investigate flaws of conventional vocabulary trees and improve the ranking results by introducing a credibility value to the TF-IDF scheme. By combining novel vocabulary trees and fast 3D geometric verification, we can improve the recall-datarate performance as well as the subjective ranking results for mobile 3D visual search.

<p>A fundamental characteristic of human visual perception is the ability to group together disparate elements in a scene and treat them as a single unit. The mechanisms by which humans create such groupings remain unknown, but grouping seems to play an important role in a wide variety of visual phenomena. I propose a neural model of grouping; through top-down control of its circuits, the model implements a grouping strategy that involves both a connection strategy (which elements to connect) and a selection strategy (spatiotemporal properties of a selection signal that segments target elements to facilitate identification). With computer simulations I explain how the circuits work and show how they can account for a wide variety of Gestalt principles of perceptual grouping. Additionally, I extend the model so that it can simulate visual search tasks. I show that when the model uses particular grouping strategies, simulated results closely match empirical results from replication experiments of three visual search tasks. In these experiments, perceptual grouping was induced by proximity and shape similarity (Palmer & Beck, 2007), by the spacing of irrelevant distractors and size similarity (Vickery, 2008), or by the proximity of dots and the proximity and shape similarity of line figures (Trick & Enns, 1997). Thus, I show that the model accounts for a variety of grouping effects and indicates which grouping strategies were likely used to promote performance in three visual search tasks. </p>

abstract: The label-feedback hypothesis (Lupyan, 2007, 2012) proposes that language modulates low- and high-level visual processing, such as priming visual object perception. Lupyan and Swingley (2012) found that repeating target names facilitates visual search, reducing response times and increasing accuracy. Hebert, Goldinger, and Walenchok (under review) used a modified design to replicate and extend this finding, and concluded that speaking modulates visual search via template integrity. The current series of experiments 1) replicated the work of Hebert et al. with audio stimuli played through headphones instead of self-directed speech, 2) examined the label feedback effect under conditions of varying object clarity, and 3) explored whether the relative prevalence of a target’s audio label might modulate the label feedback effect (as in the low prevalence effect; Wolfe, Horowitz, & Kenner, 2005). Paradigms utilized both traditional spatial visual search and repeated serial visual presentation (RSVP). Results substantiated those found in previous studies—hearing target names improved performance, even (and sometimes especially) when conditions were difficult or noisy, and the relative prevalence of a target’s audio label strongly impacted its perception. The mechanisms of the label feedback effect––namely, priming and target template integrity––are explored. / Dissertation/Thesis / Doctoral Dissertation Psychology 2019

付記する学位プログラム名: 霊長類学・ワイルドライフサイエンス・リーディング大学院 / 京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第22034号 / 理博第4538号 / 新制||理||1652(附属図書館) / 京都大学大学院理学研究科生物科学専攻 / (主査)教授 友永 雅己, 准教授 宮地 重弘, 教授 濱田 穣 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

No description available.

No description available.

No description available.

No description available.