The Synthesis of Vision and Action
Abstract
Our efforts to reconstruct the world using visual information have led to the insight that the study of Vision should not be separated from the study of a system's actions and purposes. In computational terms this relates to approaching the analysis of perceptual information processing systems through the modelling of the observer and world in a synergistic manner, not through the isolated modelling of observer and world as closed systems. The question still remains: how should such a synergistic modelling be realized? This chapter addresses the question by providing a methodology for synthesizing vision systems and integrating perception and action. In particular, we outline an architecture for purposive vision systems and present a hierarchy of navigational competences based on computational models of increasing complexity, employing representations of motion, shape, form and space. Pure computational considerations will not tell us what visual competences and representations are important to vision systems performing a set of tasks. Interaction, however, with empirical sciences such as Neurobiology, Physiology, Psychology, Ethology, etc., can give us inspiration about the visual categories relevant to systems existing in real world environments. Throughout the chapter, we describe biological findings and how they affect the choice of computational models and representations needed for the synthesis of a hierarchy of navigational competences in a working system.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1993
- Accession Number
- ADA285453
Entities
People
- Cornelia Fermueller
- Yiannis Aloimonos
Organizations
- University of Maryland