Journal of Spectral Imaging JSI—Journal of Spectral Imaging is an online, Open Access, peer-review journal publishing high-quality papers in the rapidly growing field of spectral, hyperspectral and chemical imaging. https://www.impopen.com/jsi-toc/8 IM Publications Open en © IM Publications Open LLP Journal of Spectral Imaging 2040-4565 © IM Publications Open LLP info@impopen.com Comparison of spectral selection methods in the development of classification models from visible near infrared hyperspectral imaging data https://doi.org/10.1255/jsi.2019.a4 Applications of hyperspectral imaging (HSI) to the quantitative and qualitative measurement of samples have grown widely in recent years, due mainly to the improved performance and lower cost of imaging spectroscopy instrumentation. Data sampling is a crucial yet often overlooked step in hyperspectral image analysis, which impacts the subsequent results and their interpretation. In the selection of pixel spectra for the calibration of classification models, the spatial information in HSI data can be exploited. In this paper, a variety of sampling strategies for selection of pixel spectra are presented, exemplified through five case studies. The strategies are compared in terms of the proportion of global variability captured, practicality and predictive model performance. The use of variographic analysis as a guide to the spatial segmentation prior to sampling leads to the selection of representative subsets while reducing the variation in model performance parameters over repeated random selection. Comparison of spectral selection methods in the development of classification models from visible near infrared hyperspectral imaging data

J. Spectral Imaging 8, a4 (2019). doi:10.1255/jsi.2019.a4

Applications of hyperspectral imaging (HSI) to the quantitative and qualitative measurement of samples have grown widely in recent years, due mainly to the improved performance and lower cost of imaging spectroscopy instrumentation. Data sampling is a crucial yet often overlooked step in hyperspectral image analysis, which impacts the subsequent results and their interpretation. In the selection of pixel spectra for the calibration of classification models, the spatial information in HSI data can be exploited. In this paper, a variety of sampling strategies for selection of pixel spectra are presented, exemplified through five case studies. The strategies are compared in terms of the proportion of global variability captured, practicality and predictive model performance. The use of variographic analysis as a guide to the spatial segmentation prior to sampling leads to the selection of representative subsets while reducing the variation in model performance parameters over repeated random selection.

]]>
Comparison of spectral selection methods in the development of classification models from visible near infrared hyperspectral imaging data Aoife A. Gowen Jun-Li Xu Ana Herrero-Langreo doi:10.1255/jsi.2019.a4 J. Spectral Imaging 8, a4 (2019) 2019-01-17 Journal of Spectral Imaging 2019-01-17 8 1 19 10.1255/jsi.2019.a4 https://doi.org/10.1255/jsi.2019.a4
Raman and Fourier transform infrared hyperspectral imaging to study dairy residues on different surface https://doi.org/10.1255/jsi.2019.a3 Milk is a complex emulsion of fat and water with proteins (such as caseins and whey), vitamins, minerals and lactose dissolved within. The purpose of this study is to automatically distinguish different dairy residues on substrates commonly used in the food industry using hyperspectral imaging. Fourier transform infrared (FT-IR) and Raman hyperspectral imaging were compared as candidate techniques to achieve this goal. Aluminium and stainless-steel, types 304-2B and 316-2B, were chosen as surfaces due to their widespread use in food production. Spectra of dried samples of whole, skimmed, protein, butter milk and butter were compared. The spectroscopic information collected was not only affected by the chemical signal of the milk composition, but also by surface signals, evident as baseline and multiplicative effects. In addition, the combination of the spectral information with spatial information can improve data interpretation in terms of characterising spatial variability of the selected surfaces. Raman and Fourier transform infrared hyperspectral imaging to study dairy residues on different surface

J. Spectral Imaging 8, a3 (2019). doi:10.1255/jsi.2019.a3

Milk is a complex emulsion of fat and water with proteins (such as caseins and whey), vitamins, minerals and lactose dissolved within. The purpose of this study is to automatically distinguish different dairy residues on substrates commonly used in the food industry using hyperspectral imaging. Fourier transform infrared (FT-IR) and Raman hyperspectral imaging were compared as candidate techniques to achieve this goal. Aluminium and stainless-steel, types 304-2B and 316-2B, were chosen as surfaces due to their widespread use in food production. Spectra of dried samples of whole, skimmed, protein, butter milk and butter were compared. The spectroscopic information collected was not only affected by the chemical signal of the milk composition, but also by surface signals, evident as baseline and multiplicative effects. In addition, the combination of the spectral information with spatial information can improve data interpretation in terms of characterising spatial variability of the selected surfaces.

]]>
Raman and Fourier transform infrared hyperspectral imaging to study dairy residues on different surface V. Caponigro F. Marini R. M. Dorrepaal A. Herrero-Langreo A. G.M. Scannell A. A. Gowen doi:10.1255/jsi.2019.a3 J. Spectral Imaging 8, a3 (2019) 2019-01-14 Journal of Spectral Imaging 2019-01-14 8 1 20 10.1255/jsi.2019.a3 https://doi.org/10.1255/jsi.2019.a3
Increased sensitivity in near infrared hyperspectral imaging by enhanced background noise subtraction https://doi.org/10.1255/jsi.2019.a2 Near infrared hyperspectral photoluminescence imaging of crystalline silicon wafers can reveal new knowledge on the spatial distribution and the spectral response of radiative recombination active defects in the material. The hyperspectral camera applied for this imaging technique is subject to background shot noise as well as to oscillating background noise caused by temperature fluctuations in the camera chip. Standard background noise subtraction methods do not compensate for this oscillation. Many of the defects in silicon wafers lead to photoluminescence emissions with intensities that are one order of magnitude lower than the oscillation in the background noise level. These weak signals are therefore not detected. In this work, we demonstrate an enhanced background noise subtraction scheme that accounts for temporal oscillations as well as spatial differences in the background noise. The enhanced scheme drastically increases the sensitivity of the camera and hence allows for detection of weaker signals. Thus, it may be useful to implement the method in all hyperspectral imaging applications studying weak signals. Increased sensitivity in near infrared hyperspectral imaging by enhanced background noise subtraction

J. Spectral Imaging 8, a2 (2019). doi:10.1255/jsi.2019.a2

Near infrared hyperspectral photoluminescence imaging of crystalline silicon wafers can reveal new knowledge on the spatial distribution and the spectral response of radiative recombination active defects in the material. The hyperspectral camera applied for this imaging technique is subject to background shot noise as well as to oscillating background noise caused by temperature fluctuations in the camera chip. Standard background noise subtraction methods do not compensate for this oscillation. Many of the defects in silicon wafers lead to photoluminescence emissions with intensities that are one order of magnitude lower than the oscillation in the background noise level. These weak signals are therefore not detected. In this work, we demonstrate an enhanced background noise subtraction scheme that accounts for temporal oscillations as well as spatial differences in the background noise. The enhanced scheme drastically increases the sensitivity of the camera and hence allows for detection of weaker signals. Thus, it may be useful to implement the method in all hyperspectral imaging applications studying weak signals.

]]>
Increased sensitivity in near infrared hyperspectral imaging by enhanced background noise subtraction Torbjørn Mehl Guro Marie Wyller Ingunn Burud Espen Olsen doi:10.1255/jsi.2019.a2 J. Spectral Imaging 8, a2 (2019) 2019-01-10 Journal of Spectral Imaging 2019-01-10 8 1 8 10.1255/jsi.2019.a2 https://doi.org/10.1255/jsi.2019.a2