Object Perception Mechanisms for Resolving Ambiguity

解决歧义的对象感知机制

• 批准号: 6710204
• 负责人: DANIEL J KERSTEN
• 金额: $ 34万
• 依托单位: UNIVERSITY OF MINNESOTA
• 依托单位国家: 美国
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-12-01 至 2007-11-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6710204
• 关键词: clinical research; computational neuroscience; cues; functional magnetic resonance imaging; human subject; neural information processing; psychophysiology; visual depth perception; visual perception

DESCRIPTION (provided by applicant): Humans perceive the shapes of objects quickly and reliably despite the complexity and objective ambiguities of natural images. Image features for an object are ambiguous because of the effects of projection together with large variations in occlusion, background clutter, illumination, and viewpoint. The very success of everyday vision implies neural mechanisms, yet to be understood, that organize ambiguous or "noisy" local features into objects and surfaces. Our past work in Bayesian theories of visual perception has shown how ambiguity may be resolved through the probabilistic integration of prior object knowledge with image features. We propose to extend this work to develop and test models of the functions and mechanisms underlying the resolution of ambiguity in object perception. We propose four series of experiments that use psychophysical, computational, and neuroimaging methods to investigate how the human visual system determines object shape given occlusion, constructs surface representations from ambiguous cues, estimates changes in object depth and segments shape from realistic background clutter.

描述（由申请人提供）：尽管自然图像的复杂性和客观模糊性，但人类快速且可靠地感知物体的形状。由于投影的影响以及遮挡、背景杂波、照明和视点的大变化，对象的图像特征是模糊的。日常视觉的成功意味着神经机制，尚未被理解，将模糊或“嘈杂”的局部特征组织成物体和表面。我们过去的工作在贝叶斯理论的视觉感知表明，如何通过概率集成的先验对象知识与图像特征的模糊性可能会得到解决。我们建议扩展这项工作，以开发和测试模型的功能和机制的解决对象感知的模糊性。 我们提出了四个系列的实验，使用心理物理学，计算和神经成像方法来研究人类视觉系统如何确定物体形状给定的遮挡，从模糊的线索构建表面表示，估计物体深度的变化，并从现实的背景杂波中分割形状。