Recognition of planar objects in 3D space using a modified dynamic link architecture

Author

Sim, H.C. ; Damper, R.I.

Author_Institution

Dept. of Electron. & Comput. Sci., Southampton Univ., UK

fYear

1998

fDate

4-7 Oct 1998

Firstpage

571

Abstract

Most object recognition systems are not invariant to perspective deformations. Hence, they are not suitable for general use in real world scenarios. This research aims at one of the fundamental subgoals leading to a realistic 3D object recognition system which is to detect flat objects in 3D space. In brief, the proposed system uses a new variant of neural processing known as dynamic link matching to achieve perspective invariant recognition even when the target is partially occluded in cluttered test images

Keywords

image matching; neural net architecture; object recognition; self-organising feature maps; unsupervised learning; 3D object recognition system; 3D space; cluttered test images; dynamic link matching; flat objects; modified dynamic link architecture; neural processing; object recognition systems; partially occluded target; perspective deformations; planar objects recognition; unsupervised self-organization; Computer architecture; Computer science; Deformable models; Gabor filters; Image recognition; Indexing; Neurons; Object detection; Object recognition; System testing; Target recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Image Processing, 1998. ICIP 98. Proceedings. 1998 International Conference on

Conference_Location

Chicago, IL

Print_ISBN

0-8186-8821-1

Type

conf

DOI

10.1109/ICIP.1998.727329

Filename

727329