Share Email Print

Proceedings Paper

Robust feature detection for 3D object recognition and matching
Author(s): Sharath Pankanti; Chitra Dorai; Anil K. Jain
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Salient surface features play a central role in tasks related to 3-D object recognition and matching. There is a large body of psychophysical evidence demonstrating the perceptual significance of surface features such as local minima of principal curvatures in the decomposition of objects into a hierarchy of parts. Many recognition strategies employed in machine vision also directly use features derived from surface properties for matching. Hence, it is important to develop techniques that detect surface features reliably. Our proposed scheme consists of (1) a preprocessing stage, (2) a feature detection stage, and (3) a feature integration stage. The preprocessing step selectively smoothes out noise in the depth data without degrading salient surface details and permits reliable local estimation of the surface features. The feature detection stage detects both edge-based and region-based features, of which many are derived from curvature estimates. The third stage is responsible for integrating the information provided by the individual feature detectors. This stage also completes the partial boundaries provided by the individual feature detectors, using proximity and continuity principles of Gestalt. All our algorithms use local support and, therefore, are inherently parallelizable. We demonstrate the efficacy and robustness of our approach by applying it to two diverse domains of applications: (1) segmentation of objects into volumetric primitives and (2) detection of salient contours on free-form surfaces. We have tested our algorithms on a number of real range images with varying degrees of noise and missing data due to self-occlusion. The preliminary results are very encouraging.

Paper Details

Date Published: 23 June 1993
PDF: 12 pages
Proc. SPIE 2031, Geometric Methods in Computer Vision II, (23 June 1993); doi: 10.1117/12.146640
Show Author Affiliations
Sharath Pankanti, Michigan State Univ. (United States)
Chitra Dorai, Michigan State Univ. (United States)
Anil K. Jain, Michigan State Univ. (United States)

Published in SPIE Proceedings Vol. 2031:
Geometric Methods in Computer Vision II
Baba C. Vemuri, Editor(s)

© SPIE. Terms of Use
Back to Top