Thursday, December 12, 2013

Perceptual recognition

Given a set of observations of a partially observable system and a set of categories of parts of the partially observable system it is useful for us to be able to recognize rather or not parts of the environment belong to one of our categories. There are two types of perceptual recognition: non-localized recognition and localized recognition.

The process of non-localized recognition simply compares our observations to the definition of our category in order to make a determination. Localized recognition on the other hand attempts to locate entities within our model of the external environment.

An example of localized recognition is that we given some observations of a star such as its spectra we can determine that said star is a sun-like star. On the other hand with localized recognition we can determine that said star is the sun. Localized recognition generally requires information about the location of objects in the environment relative to one another.

No comments:

Post a Comment