Robotics Research Technical Report: Three-Dimensional Object Recognition From Single Two-Dimensional Images (Classic Reprint) - Hardcover

David G. Lowe

Hardcover

ISBN 10: 0267686706 ISBN 13: 9780267686704

Publisher: Forgotten Books, 2018

Synopsis

Recognize 3D objects directly from a single 2D image without heavy depth data.

This book presents a practical approach to matching two-dimensional image features to three-dimensional object models using perceptual organization and probabilistic reasoning.

This work outlines how a vision system can identify objects from any viewpoint, even with partial data or clutter. It compares depth-based methods to a robust, depth-free framework and details how to combine bottom-up image groupings with top-down model verification for reliable recognition.

• Understand why depth reconstruction isn’t always necessary for accurate object recognition.
• Learn how perceptual organization discovers image groupings that stay stable across viewpoints.
• Explore probabilistic matching and viewpoint consistency to verify object presence.
• See practical examples of a working system that handles occlusion, noise, and complex backgrounds.

Ideal for readers seeking a practical framework for visual recognition and its real-world applications in computer vision.

Edition details

• Publisher: Forgotten Books
• Publication date: 2018
• Language: English
• ISBN 10: 0267686706
• ISBN 13: 9780267686704
• Binding: Hardcover
• Number of pages: 43