Segmentation and calibration of hyperspectral imaging for honey analysis

• Published in: Computers and electronics in agriculture Vol. 159; pp. 129 - 139
• Main Authors: Noviyanto, Ary, Abdulla, Waleed H.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01-04-2019; Elsevier BV
• Subjects: Calibration; Data processing; Honey; Honey analyses; Hyperspectral imaging; Image segmentation; Machine learning; Organic chemistry; Reflectance; Segmentation; Spectra; Statistical methods; Honey analyses; Calibration; Segmentation; Hyperspectral imaging
• ISSN: 0168-1699; 1872-7107
• DOI: 10.1016/j.compag.2019.02.006

•An efficient approach to acquire accurate and consistent honey spectra is proposed.•Manual and automatic segmentation methods accommodate changeable analysis scenarios.•A selective average function excludes outliers to extract unbiased spectra.•A calibration procedure overcomes variations in temperature and brightness.•A normalization method significantly reduces the spatial heterogeneity.
Hyperspectral imaging as a fast and non-invasive method for honey analysis has great potential to overcome the drawbacks of the conventional chemically based assessment methods. This paper discusses segmentation and calibration techniques to acquire accurate and consistent spectra in the implementation of hyperspectral imaging. These essential techniques have not yet been fully investigated despite they have significant implications over the honey analysis accuracy. Those techniques were developed and assessed under reflectance and transmittance sensing modes. The developed segmentation strategies (manual and automatic) followed by a selective average function demonstrated a reliable spectra extraction of honey samples. The developed calibration technique using a dynamic reference followed by normalisation successfully corrected the extracted spectra from distortions caused by variations in temperatures and lighting powers; also it greatly reduced the effect of spatial heterogeneity. The proposed segmentation and calibration techniques ensure the repeatability of spectral information acquisition which is very important for further processing to develop machine learning and statistically based prediction models.