• Kucheryavskiy, Sergey V.

Abstract

Hyperspectral imaging is a modern analytical technique combining benefits of digital imaging and vibrational spectroscopy. It allows to reveal and visualise spatial distribution of various chemical components. In a hyperspectral image<br/>every pixel is a spectrum (usually VNIR, SWIR or Raman) of a depicted area. Such<br/>image can be represented as a cube or a set of 2D “slices” — one slice for each spectral band. It contains large amount of data and to reveal useful information<br/>proper methods for processing and analysis are needed. Multivariate image analysis (MIA) is one of such methods widely spread among chemometicians.<br/>In most of the cases MIA treats pixels as objects, so an image cube has to be<br/>unfolded into a matrix, where rows represent pixels and columns — wavelengths.<br/>So in fact, multivariate image analysis works with an image as with a large set of<br/>spectra, without taking into account information about spatial relations of the<br/>pixels. This works well in general, especially for exploratory analysis or multivariate curve resolution, but for some specific tasks it is not beneficial at<br/>all. One of such tasks is classification or clustering of objects on hyperspectral<br/>images. An object here means a set of connected pixels, fully or partly separated<br/>from other objects. That could be, for example, tablets, cereals, biological cells,<br/>etc. If objects from opposite classes are not absolutely different (e.g. there are<br/>similar pixels)it can lead to a problem. For example, if two different tablets have<br/>the same or similar excipient and different active ingredients, some of the pixels<br/>chemically will be identical. But these similar pixels will be associated with different classes when a classification model is being calibrated. This can give<br/>unstable model and poor classification results. In the present work a classification<br/>method that combines classic image classification approach and MIA is proposed. The basic idea is to group all pixels and calculate spectral properties of<br/>the pixelgroup to be used further as a vector of predictors for calibration and<br/>class prediction. The grouping can be done with mathematical morphology methods applied to a score image where objects are well separated. In the case<br/>of small overlapping a watershed transformation can be applied to disjoint the<br/>objects. The method has been tested on several simulated andreal cases and<br/>showed good results and significant improvements in comparison with a standard<br/>MIA approach. The results as well as method details will be reported.

Topics

• impedance spectroscopy
• morphology
• clustering
• vibrational spectroscopy