A General Model of Saccadic Selectivity in Visual Search

视觉搜索中扫视选择性的通用模型

基本信息

批准号：
7305151
负责人：
Marc Pomplun
金额：
$ 20.95万
依托单位：
UNIVERSITY OF MASSACHUSETTS BOSTON
依托单位国家：
美国
项目类别：
财政年份：
2007
资助国家：
美国
起止时间：
2007-08-01 至 2009-07-31
项目状态：
已结题

项目摘要

DESCRIPTION (provided by applicant): Many of our everyday tasks, such as spotting a friend in a crowd or picking a bottle of soda from the refrigerator, require us to perform visual search. Due to this ubiquity of visual search, its study promises to shed light on the fundamental processes that control our visual attention so efficiently in natural tasks. To quantitatively assess search behavior, previous research using simple, artificial displays has employed eye-movement recording to analyze saccadic selectivity, that is, the bias of saccadic endpoints ("landing points" of eye movements) towards display items that share certain features with the search target. Recently, saccadic selectivity in natural, complex displays has been examined as well (Pomplun, 2006), giving a first insight into eye-movement control as it is performed during everyday tasks. The aim of this project is to devise, implement, and evaluate a general, computational model of saccadic selectivity in visual search tasks. Due to its quantitative nature, absence of freely adjustable parameters, and support from empirical research results, the Area Activation Model (Pomplun, Shen, & Reingold, 2003) is a promising starting point for developing such a model. Its basic assumption is that eye movements in visual search tasks tend to target display areas that provide a maximum amount of task-relevant information for processing. To advance this model towards a general model of saccadic selectivity in visual search, additional eye-movement studies are performed to provide detailed information on the influence of color and target size on saccadic selectivity. Based on the data obtained, various aspects of the influence of display and target features on eye- movement patterns are quantified. These data are used to devise the advanced version of the Area Activation Model. The crucial improvements include the elimination of required empirical a-priori information, the consideration of bottom-up activation, and the applicability of the model to search displays beyond artificial images with discrete items and features. Ideally, the resulting model will be straightforward, consistent with natural principles, and carefully avoiding any freely adjustable model parameters to qualify it as a streamlined and general approach to eye-movement control in visual search. Such a model will be important for understanding the functionality of the visual system and will also have significant impact on the fields of computer vision, human-computer interaction, and cognitive modeling. The project will deepen our understanding of visual attention in general and the visual factors underlying saccade programming in particular. Such understanding will advance the possibilities for surgical and therapeutic treatment of visual illnesses. Moreover, the results of the study can directly be applied to improve current human-computer interfaces for computer-assisted surgery and x-ray image analysis.

描述（由申请人提供）：我们的许多日常任务，例如在人群中发现朋友或从冰箱中挑选一瓶苏打水，要求我们进行视觉搜索。由于视觉搜索的这种无处不在，其研究有望阐明在自然任务中如此有效地控制视觉注意力的基本过程。为了定量评估搜索行为，先前使用简单的人造显示的研究采用了眼动记录来分析Saccadic的选择性，即SACCADIC终点的偏见（眼动物的“着陆点”）朝着与与某些功能共享某些功能的显示项目搜索目标。最近，还检查了天然，复杂显示器中的Saccadic选择性（Polplun，2006年），在日常任务中执行时，首先了解了眼动控制。该项目的目的是在视觉搜索任务中设计，实施和评估Saccadic选择性的一般计算模型。由于其定量性质，没有可自由调节的参数以及经验研究结果的支持，该区域激活模型（Pollenun，Shen和Reingold，2003年）是开发这种模型的有希望的起点。它的基本假设是，视觉搜索任务中的眼睛运动倾向于针对显示区域，这些区域可提供最大数量的与任务相关的信息进行处理。为了使该模型朝着视觉搜索中的Saccadic选择性的一般模型迈进，进行了其他眼动研究，以提供有关颜色和目标大小对Saccadic选择性影响的详细信息。根据获得的数据，量化了显示和目标特征对眼动模式的影响的各个方面。这些数据用于设计区域激活模型的高级版本。至关重要的改进包括消除所需的经验A-Priori信息，自下而上激活的考虑以及模型在具有离散项目和功能的人造图像之外搜索显示的搜索显示。理想情况下，所得模型将是直接的，与自然原理一致，并仔细避免任何可自由调节的模型参数，以便在视觉搜索中作为一种流线型和通用的眼动控制方法。这样的模型对于理解视觉系统的功能非常重要，并且还将对计算机视觉，人类计算机相互作用和认知建模领域产生重大影响。该项目将加深我们对视觉关注的理解，尤其是扫视节目的视觉因素。这种理解将促进视觉疾病的手术和治疗治疗的可能性。此外，该研究的结果可以直接应用于改善当前的人类计算机界面，以进行计算机辅助手术和X射线图像分析。