IDENTIFICATION OF PLASTICS AMONG NONPLASTICS IN MIXED WASTE BY REMOTE-SENSING NEAR-INFRARED IMAGING SPECTROSCOPY .2. MULTIVARIATE IMAGE RANK ANALYSIS FOR RAPID CLASSIFICATION

Macroscopic samples of household waste were experimentally characteriz ed by a sequence of images, taken in four distinct wavelength regions by NIRIS. The obtained three-dimensional stack of images serves as ind ividual fingerprint for each sample, A rapid data compression, followe d by an abstract factor rotation of this stack into a spectroscopicall y meaningful intermediate four-element vector by a method called multi variate image rank analysis (MIRA), finally provided a single number. This number serves as decision limit for detection of plastics among n onplastic waste. The MIRA results are independent of sample size and s ample position within the camera image. They are sensitive to only the type of sample material, MIRA was also found to be robust against ima ge errors such as shadow or slight sample replacements between measure ments.