Advanced

Natural Vision for Artificial Systems, Active Vision and Thought

Kopp, Lars LU (2003) In Lund University Cognitive Studies 100.
Abstract
The visual system in animals and humans consists of a number of information processing units that are here called visual routines. These routines are specialized in handling different cognitive tasks, for example, object recognition, motion, stereo, color, texture, and optic flow. together, these routines contribute to the internal representation of the visual world. On the top of this network of interconnected units, language is grounded. The thesis covers several fields in vision research and aims at presenting an overview of how the different fields may be connected. The thesis attempts to describe a construction of an artificial vision system that is based on what we know about natural seeing processes. The aim is to build a complete... (More)
The visual system in animals and humans consists of a number of information processing units that are here called visual routines. These routines are specialized in handling different cognitive tasks, for example, object recognition, motion, stereo, color, texture, and optic flow. together, these routines contribute to the internal representation of the visual world. On the top of this network of interconnected units, language is grounded. The thesis covers several fields in vision research and aims at presenting an overview of how the different fields may be connected. The thesis attempts to describe a construction of an artificial vision system that is based on what we know about natural seeing processes. The aim is to build a complete vision system that includes even the cognitive aspects of vision, for example, how language can exploit visual representations. According to the proposed construction, the visual process is not finished when an object centered representation is attained. Such a representation is argued to be only one among many visual aspects of the scene. Beyond a single object centered identification of an object, a focused content is also formed, which means that some aspects of the object are highlighted. A visual object is not simply constructed as a statical content of properties, but an object can always change its content. What the content will be depends on external as well as on internal factors. For example, sometimes an object A that looks similar to an object B can change so that it looks more similar to the object C. In that case the content of the focused properties has changed so that object B share more common properties with object C. Processing language requires that such a mechanism dynamically forms the internal object that is in focus. Language understanding is a process that interactively adapts the internal object in focus between the two communicators so that it will fit their internal knowledge data-bases of object categories. (Less)
Please use this url to cite or link to this publication:
author
opponent
  • Prof. Ekelundh, Jan-Olof, KTH, Stockholm
organization
publishing date
type
Thesis
publication status
published
subject
keywords
visual alphabet, representations, robotics, language, artificial vision, cognition, Artificial intelligens, Artificiell intelligens
in
Lund University Cognitive Studies
volume
100
pages
183 pages
defense location
Carolinasalen, Kungshuset, LundagÄrd, 2003 May 27th at 1 PM
defense date
2003-05-27 13:00
ISSN
1101-8453
ISBN
91-628-5631-6
language
English
LU publication?
yes
id
5563c60f-12e6-4978-9e16-48170e19da9f (old id 27604)
date added to LUP
2007-06-07 11:17:51
date last changed
2016-09-19 08:45:00
@misc{5563c60f-12e6-4978-9e16-48170e19da9f,
  abstract     = {The visual system in animals and humans consists of a number of information processing units that are here called visual routines. These routines are specialized in handling different cognitive tasks, for example, object recognition, motion, stereo, color, texture, and optic flow. together, these routines contribute to the internal representation of the visual world. On the top of this network of interconnected units, language is grounded. The thesis covers several fields in vision research and aims at presenting an overview of how the different fields may be connected. The thesis attempts to describe a construction of an artificial vision system that is based on what we know about natural seeing processes. The aim is to build a complete vision system that includes even the cognitive aspects of vision, for example, how language can exploit visual representations. According to the proposed construction, the visual process is not finished when an object centered representation is attained. Such a representation is argued to be only one among many visual aspects of the scene. Beyond a single object centered identification of an object, a focused content is also formed, which means that some aspects of the object are highlighted. A visual object is not simply constructed as a statical content of properties, but an object can always change its content. What the content will be depends on external as well as on internal factors. For example, sometimes an object A that looks similar to an object B can change so that it looks more similar to the object C. In that case the content of the focused properties has changed so that object B share more common properties with object C. Processing language requires that such a mechanism dynamically forms the internal object that is in focus. Language understanding is a process that interactively adapts the internal object in focus between the two communicators so that it will fit their internal knowledge data-bases of object categories.},
  author       = {Kopp, Lars},
  isbn         = {91-628-5631-6},
  issn         = {1101-8453},
  keyword      = {visual alphabet,representations,robotics,language,artificial vision,cognition,Artificial intelligens,Artificiell intelligens},
  language     = {eng},
  pages        = {183},
  series       = {Lund University Cognitive Studies},
  title        = {Natural Vision for Artificial Systems, Active Vision and Thought},
  volume       = {100},
  year         = {2003},
}