Presentation 2007-02-23
Accurate Image Recognition From a Small Number of Samples Using Correlation Matrix of Feature Vector : Application to Traffic Sign Recognition
Shinya TAGUCHI, Junshiro KANDA, Yoshihiro SHIMA, Jun-ichi TAKIGUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We describe a novel image recognition method using correlation matrix and apply it to a traffic sign recognition task. Such tasks are inevitable for automatic road facility maintenance and an intelligent car navigation systems, however, it is an unresolved problem to accurately recognize traffic signs under severe reflection and appearance changes. To overcome the difficulties, we propose a novel recognition method using correlation matrix. We can successfully classify 10 types of traffic signs with over 90% accuracy using a small number of sample images. Moreover, we propose a rapid object detection method using correlation coefficients and Adaboost, which detects traffic signs with over 95% accuracy under severe conditions.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Correlation Matrix / Correlation Coefficients / Adaboost / Traffic Sign / ITS
Paper # ITS2006-85,IE2006-270
Date of Issue

Conference Information
Committee ITS
Conference Date 2007/2/16(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 Intelligent Transport Systems Technology (ITS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Accurate Image Recognition From a Small Number of Samples Using Correlation Matrix of Feature Vector : Application to Traffic Sign Recognition
Sub Title (in English)
Keyword(1) Correlation Matrix
Keyword(2) Correlation Coefficients
Keyword(3) Adaboost
Keyword(4) Traffic Sign
Keyword(5) ITS
1st Author's Name Shinya TAGUCHI
1st Author's Affiliation Mitsubishi Electric Corp. Information Technology R&D Center()
2nd Author's Name Junshiro KANDA
2nd Author's Affiliation Mitsubishi Electric Corp. Information Technology R&D Center
3rd Author's Name Yoshihiro SHIMA
3rd Author's Affiliation Mitsubishi Electric Corp. Kamakura Works
4th Author's Name Jun-ichi TAKIGUCHI
4th Author's Affiliation Mitsubishi Electric Corp. Kamakura Works
Date 2007-02-23
Paper # ITS2006-85,IE2006-270
Volume (vol) vol.106
Number (no) 535
Page pp.pp.-
#Pages 6
Date of Issue