Summarising stimuli that are currently out of view

From A conversation about the brain
Jump to: navigation, search

One problem in 3D vision that exists in many spheres (e.g. playing chess) is summarising branches of a tree of possibilities. For a given optic array, approaching some portions of the array leads to a huge number of other possible visual stimuli, some lead nowhere (you hit a wall). Do I go through door A or B? Place C might be reachable through door A or B and would appear in both lists. Visual features can be labelled with a series of descriptors (contrast, orientation statistics, location relative to other features, etc). There needs to be a methods of including in this set of descriptors a summary of the all the potential consequences of approaching a region of the optic array. This seems like a tricky problem.

It is worth noting that the problem has some similarities to, but is less severe than, that of transforming an allocentric representation into a body-centred frame (e.g 'point to Paris'). In computer vision applications (Simultaneous Localisation and Mapping, SLAM), this process of relating the vehicles location and orientation to a world-based map explicitly carried out at every moment (so pointing to a distant object in the map is not a problem).