Fusion of Multiple Sensing Modalities for Machine Vision
Abstract
We report on a broad program of research in machine vision to develop an approach based upon synergistically combining diverse sensing modalities. The research projects fall into four general categories: Outdoor Scene interpretation via the Fusion of Multiple Imaging Modalities; (2) Motion Computation and Object Recognition Using Range Images; (3) Structure and Identity Based on Color and Shape Information; and (4) Autonomous Navigation. Accomplishments include the development of the AIMS (automatic interpretation using multiple sensors) knowledge-based system to interpret registered laser radar and thermal images for the detection and recognition of man-made objects in outdoor rural scenes; the development of a new approach for the detection of large man-made objects using perceptual organization techniques; new algorithms for object recognition and motion estimation, including improved algorithms for using three-dimensional (range) images to compute structure and motion; a CAD- based object recognition system; a decision-theoretical algorithm to estimate 3D structures from extended sequences of 2D images taken by a moving camera; an algorithm for matching line segments based on perceptual grouping relaxation labeling; and the construction of an autonomous mobile robot, Robo-Tex, as a testbed for navigation algorithms.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 31, 1994
- Accession Number
- ADA284762
Entities
People
- J. K. Aggarwal
Organizations
- University of Texas at Austin