RI-MINACE FILTERS TO AUGMENT SEGMENTATION OF TOUCHING OBJECTS

Authors
Citation
D. Casasent et W. Cox, RI-MINACE FILTERS TO AUGMENT SEGMENTATION OF TOUCHING OBJECTS, Pattern recognition, 31(9), 1998, pp. 1311-1317
Citations number
7
Categorie Soggetti
Computer Science Artificial Intelligence","Engineering, Eletrical & Electronic","Computer Science Artificial Intelligence
Journal title
ISSN journal
00313203
Volume
31
Issue
9
Year of publication
1998
Pages
1311 - 1317
Database
ISI
SICI code
0031-3203(1998)31:9<1311:RFTASO>2.0.ZU;2-A
Abstract
RI-MINACE filters are described for a rotation-invariant distortion-in variant multi-object detection problem. The objects are agricultural p roducts, pistachio nuts, and the sensor is X-ray. The nuts can have an y orientation hence rotation invariance is necessary. Distortion invar iance is needed because each nut varies in size and shape. To detect t he separate correlation peaks, we use peaksorting with a window and mi nimum threshold. Design issues that arise in this application include: use of gray scale or binary images; the circular harmonic function or der to use; selection of the reference object, the size of the filter, the MINACE control parameter c, and the peaksorting parameters. (C) 1 998 Pattern Recognition Society. Published by Elsevier Science Ltd. Al l rights reserved.