The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

Research Paper #621

Title:Acquisition and Interpretation of 3-D in Machine Vision: Imagine
Authors:Fisher,RB; Trucco,E; Fitzgibbon,AW; Waite,M; Orr,M
Date:Apr 1993
Presented:Accepted for publication by SENSOR REVIEW, April 93
Abstract:Range data have become increasingly popular in recent years. Range sensors can acquire shape directly, thus avoiding the difficulties of methods using multiple intensity images, and have been adopted in many applications: bin picking, robotic assembly, inspection, recognition, robot navigation, automated cartography and medical diagnosis to name but a few. The Machine Vision Group (MVG) of the Department of Artificial Intelligence, University of Edinburgh, has been investigating advanced techniques for applied range image understanding for ten years, in the framework of several projects funded by British and European sources (e.g. Alvey, SERC, and the European Institute of Technology). One of the most recent actions is LAIRD (Location and Inspection with Range Data), a collaboration between British Aerospace, BAeSEMA Ltd, NEL, Heriot-Watt University, and the Universities of Edinburgh and Surrey, funded by the Information Engineering Directorate. The consortium has recently demonstrated an automated inspection system for 3-D small industrial manufacts based on depth data. MVG contributed modules for model matching and vision-oriented solid modeling. the main focus of MVG's research has been the design of a 3-D vision system, IMAGINE, for the recognition and location of complex objects from range date [1]. Several prototypes have been assembled in the course of the years and a number of spin-off projects carried out in various areas of machine vision, including stereo and motion analysis, automatic determination of grasping points, and MRI image analysis. This paper describes IMAGINE developments in data acquisition, surface segmentation, 3-D model selection and matching, solid modeling, as well as a recently started research effort on automatic model acquisition.

[Search These Pages] [DAI Home Page] [Comment]