Presentation 1997/12/18
A recognition method of 3-D object arrangement using optimization of global similarity to acquired image
Manabu Hashimoto, Akinobu Seki, Kazuhiko Sumi, Shin'ichi Kuroda, Tetsuji Haga,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we propose a scene recognition method for stacked hexahedrons. A conventional technique like a template matching can not recognize multiple objects using the similarity between the model and each object(local similarity), because there are many false patterns on the objects. If the scene is completely interpreted, not only local similarity but also the similarity of assumed image to acquired image(global similarity) will become high. A method we proposed is based on maxmizing this global similarity. It extracts the object candidates from the image, generates a lot of hypotheses as the combination of candidates, and then a hypothesis which has highest global similarity is chosen. We also applied the genetic algorithm to the hypothesis generation/verification process. Through real image experiments, we show that our method is practical.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) scene interpretation / global similarity / hypothesizing and verification / genetic algorithm
Paper # PRMU97-177
Date of Issue

Conference Information
Committee PRMU
Conference Date 1997/12/18(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A recognition method of 3-D object arrangement using optimization of global similarity to acquired image
Sub Title (in English)
Keyword(1) scene interpretation
Keyword(2) global similarity
Keyword(3) hypothesizing and verification
Keyword(4) genetic algorithm
1st Author's Name Manabu Hashimoto
1st Author's Affiliation Industrial Electronics and Systems Laboratory Mitsubishi Electric Corporation()
2nd Author's Name Akinobu Seki
2nd Author's Affiliation Industrial Electronics and Systems Laboratory Mitsubishi Electric Corporation
3rd Author's Name Kazuhiko Sumi
3rd Author's Affiliation Industrial Electronics and Systems Laboratory Mitsubishi Electric Corporation
4th Author's Name Shin'ichi Kuroda
4th Author's Affiliation Industrial Electronics and Systems Laboratory Mitsubishi Electric Corporation
5th Author's Name Tetsuji Haga
5th Author's Affiliation Industrial Electronics and Systems Laboratory Mitsubishi Electric Corporation
Date 1997/12/18
Paper # PRMU97-177
Volume (vol) vol.97
Number (no) 458
Page pp.pp.-
#Pages 8
Date of Issue