NIR Detection of pits in cherries

Authors: Haff, Ronald - Ron; Liang, Peishih

Submitted to: Proceedings of the American Society of Agricultural and Biological Engineers International (ASABE)
Publication Type: Abstract Only
Publication Acceptance Date: 10/1/2016
Publication Date: 11/11/2016
Citation: Haff, R.P., Liang, P. 2016. NIR Detection of pits in cherries. Proceedings of the American Society of Agricultural and Biological Engineers International (ASABE). November 6-11,2016. lb.

Interpretive Summary: Spectra from the short wavelength NIR region (700-1100 nm) are thought to provide better results when used to analyze high moisture agricultural products than do spectra at longer wavelengths (1100-2500 nm). This is because spectrometers have for the most part employed gratings, which limit the light intensity available for analysis. Thus, the shorter wavelengths with higher energy and sample penetration depths provide better results. In this study cherry sugar concentration measurements were used as a model for comparing accuracy of results between the short and long wavelength NIR regions using a spectrometer based on the Michelson-based Fourier Transform (FT) configuration, with the hypothesis being that the FT design eliminates the limitations as regards light intensity inherent in grating based instruments. Spectra of fresh cherries were acquired under two conditions: (i) 16 cm-1 resolution and 64 scans (short wavelength region; the standard setting for agriculture products) and; (ii) 64 cm-1 resolution and 128 scans (long wavelength region). The results indicate that while spectra acquired under the typical setting of 16 cm-1 resolution provide excellent calibration results (RMSEP 0.41 Brix and RPD 4.46), the 64 cm-1 setting yielded a better calibration with a RMSEP 0.38 Brix and an RPD of 4.87, supporting the hypothesis that with the FT instrument the long wavelength region provides equal or better spectra for calibrations when analyzing high moisture agricultural commodities.

Technical Abstract: It is generally accepted that spectra from the short wavelength NIR region (700-1100 nm) provide a superior basis for analysis of high moisture agricultural products than do spectra at longer wavelengths (1100-2500 nm). This is because spectrometers have for the most part employed gratings, which limit the light intensity available for analysis. Thus, the shorter wavelengths with higher energy and sample penetration depths provide better results. In this study cherry brix measurements were used as a model for comparing accuracy of results between the short and long wavelength NIR regions using a spectrometer based on the Michelson-based Fourier Transform (FT) configuration, with the hypothesis being that the FT design eliminates the limitations as regards light intensity inherent in grating based instruments. Spectra of fresh cherries were acquired under two conditions: (i) 16 cm-1 resolution and 64 scans (short wavelength region; the standard setting for agriculture products) and; (ii) 64 cm-1 resolution and 128 scans (long wavelength region). The results indicate that while spectra acquired under the typical setting of 16 cm-1 resolution provide excellent calibration results (RMSEP 0.41 'Brix and RPD 4.46), the 64 cm-1 setting yielded a better calibration with a RMSEP 0.38 'Brix and an RPD of 4.87, supporting the hypothesis that with the FT instrument the long wavelength region provides equal or better spectra for calibrations when analyzing high moisture agricultural commodities.