Near-infrared reflectance analysis for prediction of cooked rice texture

The ability of near-infrared (NIR) spectroscopy to predict sensory texture attributes of diverse rice cultivars was examined. The sensory texture of 8 7 samples representing 77 different short-, medium-, and long-grain cultiva rs was evaluated by trained panelists using descriptive analysis. Correlati ons between sensory texture attributes and NIR reflectance data were examin ed using the multivariate method of partial least squares (PLS) regression. Texture attributes (hardness, initial starchy coating, cohesiveness of mas s, slickness, and stickiness) measured by panelists in the early evaluation phases were successfully predicted (R-calibration(2) 0.71-0.96). Cohesiven ess of mass, the maximum degree to which the sample holds together in a mas s while chewing, was best modeled with R-calibration(2) = 0.96 and R-valida tion(2) = 0.90. Key wavelengths contributing to the models describing the t exture attributes were wavelengths also contributing to models for amylose, protein, and Lipid contents.