Computational Modeling and Analysis of Human Gaze Behavior

计算机视觉系统越来越多地使用焦点处理和定向凝视来处理现实环境中的复杂问题，如对象识别。通过理解这个问题领域，还可以以原则性的方式解决计算机科学的其他一些重要领域，例如设计用户界面、引导机器人查看相关项目、增强数据的可视化，或者理解为什么有人可能会在驾驶时被车内显示器分心。 从历史上看，技术限制将对这些问题的研究限制在相对有限的场景中，例如在实验室的计算机显示器上查看图片。最近的技术进步为走出实验室进入现实世界提供了可能性。这项研究的目标是一组丰富的逼真的观看场景，范围从考虑3D，到以自我为中心的观看，再到在存在环境的完整结构表示的情况下进行建模。 这包括在计算水平上理解许多影响凝视分配的因素，包括语境在观看行为中所起的作用。如果有人在森林里看到明亮的辉光，这可能是一个相当大的事件，但同样的灯光在时代广场中央闪烁，不值得重新考虑。 提出的新的研究路线将极大地丰富观看行为的计算模型及其与广泛的重要应用领域的联系。解决这些问题的性质将在激励进一步努力方面产生影响，这些努力在解决这些问题时采用类似的形式主义，同时在国际一级具有广泛的影响力和知名度。

Computer vision systems are increasingly using focal processing, and directed gaze to deal with complex problems such as object recognition in real-world environments. In understanding this problem domain, it also becomes possible to address a number of other important areas of Computer Science in a principled manner such as designing a user interface, guiding robots to look at relevant items, enhancing visualization of data, or understanding why someone might become distracted by an in car display while driving. Historically technological limitations confined the study of these issues to relatively constrained scenarios such as viewing pictures on a computer monitor in a lab. More recent technological advancements present the possibility of moving outside of the lab and into the real world. This research targets a rich set of realistic viewing scenarios that range from considering 3D, to egocentric viewing, to modeling in the presence of a complete structural representation of the environment. This includes understanding a number of contributing factors to gaze allocation at a computational level including the role that context plays in viewing behavior. If one were to spot a bright glowing light in the forest, this may be quite an event, but the same light flashing in the middle of Times Square wouldn't warrant a second thought. The novel lines of research put forth will serve to significantly enrich the body of computational models of viewing behavior and their ties to a wide range of important application domains. The nature in which these problems are addressed will have impact at the level of inspiring further efforts that apply a similar formalism in addressing these issues, while having broad influence and visibility at an international level.