Data Mining from Near-infrared Spectra of Plastics by 1/f Fluctuation

Mitsue ONODERa, Umpei NAGASHIMAb*, Sumie KATOc, Haruo HOSOYAa, Narushi GOTOd, Toshio AMANOe, Kazutoshi TANABEb and Hiroyuki UESAKAf

aDepartment of Information Sciences, Ochanomizu Univ.,
2-1-1 Otsuka, Bunkyo-ku, Tokyo 112-8610, Japan.
bNational Institute of Materials and Chemical Research, MITI,
1-1 Azuma, Tsukuba 305-8565, Japan.
cGraduate School of Humanities and Sciences, Ochanomizu Univ.,
2-1-1 Otsuka, Bunkyo-ku, Tokyo 112-8610, Japan.
dMolecular Simulation Inc.,
2-17-8 Hamacho, Nihonbashi, Chuo-ku, Tokyo 103-0007, Japan.
eOpt Reserch Inc.,
3-1-6 Nihonbashi, Chuo-ku, Tokyo 103-0027, Japan.
fFujitsu Ltd.,
Takezono, Tsukuba 305-0032, Japan.

(Received: January 5, 1998 ; Accepted for publication: March 19, 1998 ; Published on Web: June 16, 1999)

A rapid and intact method was developed for discriminating plastics tips by the analysis of 1/f fluctuation of near-infrared reflection spectra. First, the regression lines of the power spectrum of a given sample are obtained after middle pass filtering with FFT. Then the data point of the cross section and slope of the obtained regression line are plotted. According to our test measurement, eighteen samples were clearly classified into five groups of different chemical compositions. This novel method was found to be useful for recycling plastics waste efficiently and with reasonable accuracy.

Keywords: FFT, Plastic waste, Near-infrared spectrum, Data mining, 1/f fluctuation

Abstract in Japanese

Text in Japanese

PDF file(274KB)

Author Index
Keyword Index