Mechanisms of spatial localization in human vision

点击翻译按钮获取中文摘要

How do you determine where a car is when you're about to cross a busy street? Introspectively, this seems effortless and self-evident, but localizing an object in the world is far from simple. Determining the car's location, for example, requires taking into account eye position, head position, how the car is moving and where it is relative to other parts of the scene (such as the crosswalk). All this must be done accurately, as even a small error in determining the location of a car could lead to a devastating crash. Our ability to interact with the world depends on how we determine an object's location, yet we lack a complete description of how we do this. The projects in this proposal are aimed at developing a testable model that accounts for the perceived locations of objects in these types of complex environments, based on two components. The first is retinal image content (what we see) - this includes taking into account interactions with the motion of the object, the positions of other objects now, as well as where we're looking now and have been looking recently. The other component is individual variability in perceived position. Even in the simplest of situations, our ability to localize and interact with objects in the world varies considerably between individuals and across different locations in the visual field. When making difficult perceptual judgments of location, participants consistently and idiosyncratically misjudge the locations of objects, exhibiting `signatures' or fingerprints of error at different locations. In order to develop this model, it is essential to first understand exactly how this individual variability contributes to perceived location. Therefore, to move towards this long-term goal, this proposal asks: (1) whether our individual spatial representations fixed, or do they change over time, (2) at what stage of visual processing do they impact localization, and (3) how we combine our individual maps with information from the world. In addition to informing fundamental visual processes, a comprehensive model of visual localization is essential to develop technologies that support perceptual decisions in complex environments and make assumptions about where humans perceive objects in the world, including autonomous driving systems, computer-aided diagnosis in radiology, and AI-based guidance systems. There are a host of technologies that exist to help us in the world, and all of them need to have some understanding of the gap between where a computer and the user might localize an object. The proposed work will also provide a rich source of training opportunities for students, in experiment design, analysis and communication skills. This will prepare students for a range of careers (e.g., natural sciences, health sciences, engineering, the private sector or in government), and will have a positive impact on the development of Canada's future work force.