Digital Image Analysis Method for Rapid Measurement of Rice Degree of Milling

Cereal Chem. 75(3):380-385 A digital image analysis method was developed to quickly and accurately measure the degree of milling (DOM) of rice. The digital image analysis method was statistically compared to a chemical analysis method for evaluating DOM, which consisted of measuring the surface lipids concentration (SLC) of milled rice. The surface lipid area percentage (SLAP) obtained by the image analysis method and the SLC obtained by chemical analysis had a high coefficient of determination using a quadratic model (R = 0.9819) and using a logarithmic model (R = 0.9703). The quadratic model and the logarithmic model were validated using the test data set and it received high coefficients of determination (R = 0.9502 and R = 0.9459, respectively). Rice millers grade rice quality for nutritional and economic considerations and, thus, need a fast and accurate grading system as part of their milling assessment operations. The primary physical grading factors are head rice yield, which is the weight percentage of rough rice remaining as head rice (kernels that are 75% or more of their original length after milling), and degree of milling (DOM), which indicates how much bran remains on milled kernels (USDA 1979). Generally, milled rice with a high DOM level has less kernel bran than does milled rice with low DOM levels. The current methods for grading rice quality are subjective and time-consuming. This study developed a digital image analysis method that can quickly and accurately determine DOM. DOM is an important factor in terms of the nutritional value and the economic return of the milled rice. Low DOM level rice contains more protein, vitamins, minerals, and lipids than does high DOM rice (Wadsworth et al 1991). Although low DOM level rice has greater nutritional value, it often has a lower market appeal because most consumers prefer the taste and appearance of well-milled rice. Additionally, the degree to which rice is milled inversely affects head rice yield (Sun and Siebenmorgen 1993). Therefore, adjusting DOM during the rice milling operations is essential for optimizing quality and economic return. DOM can be measured by several methods, including visual inspection, chemical analysis, and optical measurements. Traditionally, DOM has been determined through visual inspection by trained personnel. For official grading, this judgment is made by comparing a sample to one of four official samples representing the four DOM grades (undermilled, lightly milled, reasonably well-milled, and well-milled) defined by the United States Standards for Milled Rice (USDA 1979). The closest similarity between the official representative sample and the inspection sample determines the DOM grade. Visual inspection is not only subjective but also is lacking in terms of quantitatively assessing the milling degree. For accurate measurement, more objective and quantitative methods must be employed to determine DOM. Chemical methods of assessing DOM include the differential dye-staining procedure and the compositional analysis procedure. The differential dye-staining procedure augments the visual difference between endosperm and bran, whereas the compositional analysis method quantifies the amount of bran or bran constituents that remain on the rice kernel. Desikachar (1955), Borasio (1955), Bhattacharya and Sowbhagya (1972), and FAO (1972) used the differential dye-staining method to estimate rice DOM. Although the differential dye-staining method more readily distinguishes the difference between endosperm and bran, the assignment of DOM is still somewhat subjective because this method requires a visual assessment rather than a quantitative measurement. Barber and Benedito (1976) further evolved the differential dye-staining procedure into a quantitative procedure by defining the color bran balance (CBB) index. Milled rice kernels were soaked with a methylene blue and eosine solution in methyl alcohol, staining the bran area blue and the endosperm pink. The total bran and kernel areas were measured by planimetry. A CBB index was then assigned to indicate the DOM based on the planimetry measurements. Several other researchers attempted to determine DOM by analyzing the constituents that remained on the kernel after milling, primarily surface lipids (Autry et al 1955, Hogan and Deobald 1961, Watson et al 1975, Matthews and Spadaro 1980, Siebenmorgen and Sun 1994, Chen and Siebenmorgen 1997). However, all these chemical methods take hours to get the DOM data. Therefore, these chemical methods are too time-consuming for routine DOM measurement. Optical measurement of DOM is contingent upon the reflection of light or the transmission of light through milled rice at selected wavelengths. Kik (1951), Stermer et al (1968), Johnson (1965), Siebenmorgen and Sun (1994), Archer and Siebenmorgen (1995) have reported the use of optical instruments for measuring DOM. Kao (1986) and Wadsworth et al (1991) developed near-infrared spectroscopy procedures to ascertain DOM. Because all these instruments utilized a bulk of rice kernels to estimate DOM, it was impossible to determine the surface lipid distribution on individual kernels. By using machine vision, each single rice kernel can be scanned completely, and DOM information can be obtained accurately through the image of each kernel. Machine vision and image processing techniques have been widely applied throughout the agricultural and food processing fields, particularly in the quality inspection and sorting of food materials. Machine vision techniques provide a quick and objective means for measuring or evaluating the visual features of products. Researchers reported using these techniques for peach defect detection (Miller and Delwiche 1991), potato inspection (Tao et al 1990), fruit sorting (Tao 1996a,b; Tao et al 1995a,b), apple bruise detection (Upchurch et al 1994; Throop et al 1995; Tao 1996a,b; Wen and Tao 1998), corn kernel breakage classification (Liao et al 1993), corn kernel stress crack detection (Yie et al 1993), wheat classification (Zayas et al 1996, Zayas and Steele 1996), and grain classification (Shatadal et al 1995). However, there are few reports on using machine vision to measure DOM of rice. Fant et al (1994) discussed using gray-scale inten1 Published with the approval of the Director of the Agricultural Experiment Station, University of Arkansas, Fayetteville, AR. Mention of trademark or proprietary products does not constitute a guarantee or warranty by the University of Arkansas and does not imply its approval to the exclusion of other products that may also be suitable. 2 Former graduate assistant, assistant professor, professor, and research assistant, respectively, Biological and Agricultural Engineering Department, University of Arkansas, Fayetteville, AR. 3 Corresponding author. E-mail: [email protected] Publication no. C-1998-0408-05R. © 1998 American Association of Cereal Chemists, Inc.