Skip to main content
ARS Home » Plains Area » Manhattan, Kansas » Center for Grain and Animal Health Research » Stored Product Insect and Engineering Research » Research » Publications at this Location » Publication #363405

Research Project: Impacting Quality through Preservation, Enhancement, and Measurement of Grain and Plant Traits

Location: Stored Product Insect and Engineering Research

Title: Detection of chalk in single kernels of long-grain milled rice using imaging and visible/near infrared instruments

Author
item Armstrong, Paul
item McClung, Anna
item Maghirang, Elizabeth
item Chen, Ming Hsuan
item Brabec, Daniel - Dan
item YAPTENCO, K - University Of The Philippines Los Banos
item FAMOSO, A - Louisiana State University
item ADDISON, C - Louisiana State University

Submitted to: Cereal Chemistry
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/8/2019
Publication Date: 9/14/2019
Citation: Armstrong, P.R., McClung, A.M., Maghirang, E.B., Chen, M., Brabec, D.L., Yaptenco, K.F., Famoso, A.N., Addison, C.K. 2019. Detection of chalk in single kernels of long-grain milled rice using imaging and visible/near infrared instruments. Cereal Chemistry. 96(6):1103-1111. https://doi.org/10.1002/cche.10220.
DOI: https://doi.org/10.1002/cche.10220

Interpretive Summary: To maintain the competitiveness of U.S. long grain rice in U.S. and foreign markets, having translucent whole milled grain is critical. A good instrumented method to detect Chalky grain will provide breeders and industry with an effective tool for developing low-chalk varieties or growing practices that reduce chalk occurrence. Two instruments developed at the Center for Grain and Animal Health Research, U.S. Department of Agriculture-Agricultural Research Service (USDA-ARS), a single kernel near-infrared (SKNIR) instrument and a light-emitting diode (SiLED) high speed sorter, were compared with two commercially-available imaging instruments, WinSEEDLE and SeedCount used for chalk measurement. Three, two-way chalk classifications were defined for single kernels based on visual inspection: (1) <50% or =50% opacity or chalk (modified-GIPSA), (2) <10% or =10% opacity (10% Cut-off) and (3) 100 % opacity or 100% translucent (MaxLevel). The SKNIR method provided the best classification for the modified-GIPSA definition with an 82.4% average correct classification (CC). The WinSEEDLE had the best classification for the 10% Cut-off definition, with an 84% CC for nonchalky kernels and a 96% CC for chalky kernels. For the MaxLevel definition, average CCs of both the SKNIR and SiLED methods were similar, at 93% and 95%, respectively. The average CCs were much lower for the WinSEEDLE and SeedCount method at 14% and 58%, respectively. The low values are a result of using an absolute threshold of 100% for chalky or nonchalky kernels, where a very small misclassification of the kernel image will cause error. All the instruments can be used to classify chalk, but their accuracy depends the chalk definition used.

Technical Abstract: To maintain the competitiveness of U.S. long grain rice in U.S. and foreign markets, having translucent whole milled grain is critical. An objective technique to detect grain chalk, opaque areas in the grain, will provide breeders and industry with an effective tool for developing low-chalk varieties or agronomic practices that reduce chalk occurrence. Two instruments developed at the Center for Grain and Animal Health Research, U.S. Department of Agriculture-Agricultural Research Service (USDA-ARS), a single kernel near-infrared (SKNIR) tube instrument and a silicon-based light-emitting diode (SiLED) high speed sorter, were compared with two commercially-available imaging instruments, WinSEEDLE and SeedCount used for chalk quantification. Three, two-way chalk classifications were defined for single kernels based on visual inspection: (1) <50% or =50% opacity or chalk (modified-GIPSA), (2) <10% or =10% opacity (10% Cut-off) and (3) 100 % opacity or 100% translucent (MaxLevel). The SKNIR method provided the best classification for the modified-GIPSA definition with an 82.4% average correct classification (CC), i.e., 89% and 76% for nonchalky and chalky kernels, respectively. The WinSEEDLE had the best classification for the 10% Cut-off definition, with an 84% CC for nonchalky kernels and a 96% CC for chalky kernels. For the MaxLevel definition, average CCs of both the SKNIR and SiLED methods were similar, at 93% and 95%, respectively. The average CCs were lower for both the WinSEEDLE method and the SeedCount method at 14% and 58%, respectively. These low CC values are a result of using a threshold of 100% for chalky or nonchalky kernels, where a single misclassified pixel within the image will cause misclassification. Calibration models developed for both the SKNIR and SiLED methods indicate that their classifications were based mainly on spectral differences near the adsorption bands for starch, protein and water content. All the instruments can be used to classify chalk, but their level of accuracy depends on how chalk is defined.