Share Email Print

Proceedings Paper

Volumetric segmentation of range images for printed circuit board inspection
Author(s): Erik R. Van Dop; Paul P. L. Regtien
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Conventional computer vision approaches towards object recognition and pose estimation employ 2D grey-value or color imaging. As a consequence these images contain information about projections of a 3D scene only. The subsequent image processing will then be difficult, because the object coordinates are represented with just image coordinates. Only complicated low-level vision modules like depth from stereo or depth from shading can recover some of the surface geometry of the scene. Recent advances in fast range imaging have however paved the way towards 3D computer vision, since range data of the scene can now be obtained with sufficient accuracy and speed for object recognition and pose estimation purposes. This article proposes the coded-light range-imaging method together with superquadric segmentation to approach this task. Superquadric segments are volumetric primitives that describe global object properties with 5 parameters, which provide the main features for object recognition. Besides, the principle axes of a superquadric segment determine the phase of an object in the scene. The volumetric segmentation of a range image can be used to detect missing, false or badly placed components on assembled printed circuit boards. Furthermore, this approach will be useful to recognize and extract valuable or toxic electronic components on printed circuit boards scrap that currently burden the environment during electronic waste processing. Results on synthetic range images with errors constructed according to a verified noise model illustrate the capabilities of this approach.

Paper Details

Date Published: 3 October 1996
PDF: 8 pages
Proc. SPIE 2899, Automated Optical Inspection for Industry, (3 October 1996); doi: 10.1117/12.252997
Show Author Affiliations
Erik R. Van Dop, Univ. Twente (Netherlands)
Paul P. L. Regtien, Univ. Twente (Netherlands)

Published in SPIE Proceedings Vol. 2899:
Automated Optical Inspection for Industry
Frederick Y. Wu; Shenghua Ye, Editor(s)

© SPIE. Terms of Use
Back to Top