Presentation 2004/9/3
Data Learning for Color Target Detection with Nearest Neighbor Classifier in the Posit ion-Color Space
Norimichi Ukita,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We have proposed a method for detecting foreground objects in non-stationary scenes[8]. With this method, rea-time detection can be realized by employing nearest neighbor classifier with xy-YUV 5D space, consisting of the x and y coordinates of an image and Y, U and V levels of a color. With the rules described in [8], however, the performance of the proposed method may not be demonstrated because 1) it is difficult to learn all variations of a background scene in advance and 2) the thresholds for learning and detecting target colors are fixed. In this paper, therefore, we 1) generate the background data from observed background colors with interpolation and outlier elimination taking into account the characteristics of the YUV color space and 2) determine the threshlods for learning and detecting target colors depending on the variances of observed background colors in each pixel. We conducted experiments to confirm the effectiveness of the method for improving the performance of object detection.
Keyword(in Japanese) (See Japanese page)
Keyword(in English)
Paper # PRMU2004-64
Date of Issue

Conference Information
Committee PRMU
Conference Date 2004/9/3(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) Data Learning for Color Target Detection with Nearest Neighbor Classifier in the Posit ion-Color Space
Sub Title (in English)
Keyword(1)
1st Author's Name Norimichi Ukita
1st Author's Affiliation Graduate School of Information Science, Nara Institute of Science and Techonology:PRESTO, JST()
Date 2004/9/3
Paper # PRMU2004-64
Volume (vol) vol.104
Number (no) 290
Page pp.pp.-
#Pages 8
Date of Issue