Features and Objects in the Visual System

视觉系统分为专门用于物体识别的腹侧流和专门用于运动处理和视觉运动指导（如抓取物体）的背侧流。由于这种功能分离，在腹侧流中有许多物体表征模型，在背侧流中有许多运动处理模型，但研究大多缺乏对背侧流中物体表征的性质或存在的调查。基于功能上的差异，这两个流将以不同的方式使用这些对象，因此也以不同的方式表示它们。也就是说，腹侧流将处理与物体识别相关的特征，包括形状、颜色、纹理等。而背流则处理与动作相关的特征，如空间定位、运动、抓握的三维形状等。因此，正确地将相关特性集成或绑定到每个流中变得非常重要。这两个流都处理特征，这些特征选择神经元需要连接或集成到一个分布式网络中，以形成对象的表示。特征集成，又称绑定问题，是神经科学中一个长期存在的突出问题。本提案中的项目研究了背流中特征集成和对象表示的本质。它们将决定哪些特征被整合到背流中，这些特征如何被整合到背流对象表征中，背流中依赖于特征的处理阶段与那些对对象表征起作用的处理阶段，以及对象表征是如何在背流中被选择和处理的。提案中的研究很重要，因为他们的结果将为背流处理产生新的可测试模型，并测试和约束当前的特征集成和对象表示模型。这项研究对视觉科学家、认知和系统神经科学家、计算建模和计算机视觉研究人员都很重要，这些都是加拿大的优势领域。

The visual system is divided into a ventral stream specialized for object recognition and a dorsal stream specialized for motion processing and visuomotor guidance such as grasping objects. Because of this functional separation, there are many models of object representation in the ventral stream and many models of motion processing in the dorsal stream, but research has been mostly devoid of investigations as to the nature or existence of object representations in the dorsal stream. Based on the difference in functionality, the two streams would use these objects differently and therefore represent them differently too. That is, the ventral stream would be processing features relevant for object recognition, including shape, color, texture, etc. While the dorsal stream processes features relevant for action, such as spatial location, motion, 3d shape for grasping, etc. Thus, it becomes very important to correctly integrate or bind' the relevant features into each stream. Both streams process features, and these feature-selective neurons need to be connected or integrated into a distributed network to form a representation of the object. Feature integration, aka the Binding Problem, is a long outstanding problem in neuroscience.The projects in this proposal investigate the nature of feature integration and object representations in the dorsal stream. They will determine which features are integrated into the dorsal stream, how those features are integrated into a dorsal stream object representation, the stages of processing in the dorsal stream that are dependent on features versus those that work on object representations, and how object representations are selected and attended to within the dorsal stream. The studies in the proposal are important in that their results will produce new testable models for dorsal stream processing as well as test and constrain current models of feature integration and object representation. This research is important to vision scientists, cognitive and systems neuroscientists, computational modeling and computer vision researchers, which are all areas of strength in Canada.