A Multi-Classifier Approach to Modelling Human and Automatic Visual Cognition

Computer vision is afield which addresses many of the functional characteristics commonly associated with human vision. For example, identifying objects in a complex scene is a typical - and difficult - problem, but represents a task domain which well illustrates the way in which insights at the human-machine interface can be mutually beneficial, and is the area on which this paper focuses. Specifically, there is great current security interest in recognising human faces, and this task provides a very typical and important context for the system proposed though our system is also concerned with the study of less complex objects. The system seeks to develop working models of the operation of the human visual cognition system via a comparison between empirical experimentation on human subjects and the construction of an automated device to mimic the results of the human experimentation based on the operation of Multi- classifier systems (MCS).