Click on “Download PDF” for the PDF version or on the title for the HTML version.


If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Measurement of Degree of Milling for Rice Using Hyperspectral Imaging

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  Paper number  131594988,  2013 Kansas City, Missouri, July 21 - July 24, 2013. (doi: http://dx.doi.org/10.13031/aim.20131594988) @2013
Authors:   Wei-Tung Chen, Yan-Fu Kuo
Keywords:   Degree of milling Rice lipid content Hyperspectral imaging Machine learning Support vector machine

Abstract. This work proposes a method to measure degree of milling for rice using hyperspectral imaging (HSI). The degree of milling (DOM) for rice is directly related to lipid content on its surface. The DOM decreases as the lipid content on rice surface decreases. HSI is a sensing technique that combines both the spatial and spectral information in the visible and near-infrared wavelength region. It is often applied for chemical compound detection. In this work, the HSI is applied to non-destructively detect rice surface lipid for DOM estimation. In the experiment, the TK Number 9 rice cultivated in Taiwan and harvested in 2009 was selected. The rice was milled using 2-kg weight loading for 60s. The samples were scanned using the HSI system in the wavelength range of 400–1000 nm at 4.7nm intervals. After that, the rice samples were dyed with Sudan black, so that the lipid residual on rice surface could be identified using regular optical microscopy and image processing algorithms. A support vector machine classifier was then developed with machine learning algorithms. The classifier can predict lipid residual on the surface of rice with the reflectance intensities between the wavelengths of 485 and 930nm by using the HS images as inputs. The model-predicted lipid image was compared pixel by pixel to a microscopic lipid image to evaluate the model performance. It was demonstrated that the developed model can predict the residual lipid distribution on rice surface at an accuracy of 85.0%.

(Download PDF)    (Export to EndNotes)