You are not logged in You are not logged in to this journal. Log In

LOG IN or SELECT A PURCHASE OPTION:

Biclops: An Active Vision System

Space 2002 and Robotics 2002
Proceedings of Space 2002: the Eighth International Conference and Robotics 2002: The Fifth International Conference
Krishna Jonnalagadda1, Ron Lumia2, and Greg Starr2

1Graduate Student, Mechanical Engineering, University of New Mexico
2Professor, Mechanical Engineering, University of New Mexico

  • Abstract
The paper presents a classification based active search method to extract 3D information about objects in the workspace. The search requires object features, which remain static with observer motion. We propose a two‐step visual exploration method for constructing 3D objects with minimum number of viewpoint placements for the vision system. The first step involves classification of objects into categories such as cylinders and boxes based on local geometric features. In the second step an active search uses local geometrical features specific to the classification to compute new viewpoints. After a sufficient number of views, the 3D object model is created. This approach results in a 3D model that is both accurate and computationally rapid.

© 2006 ASCE

ARTICLE DATA

PUBLICATION DATA

ISBN:

0-7844-0625-1

Publisher


Close

close