Chalkiness in Rice: Potential for Evaluation with Image Analysis

Chalkiness is a major concern in rice ( Oryza sativa L.) breeding because it is one of the key factors in determining quality and price. Evaluation of chalkiness is traditionally performed by human visual inspection, and there is no standard objective method to effectively classify chalky grains into different categories. In this study, we evaluated the effectiveness of image information processing with an inexpensive personal computer and a digital image scanner to measure and categorize chalkiness and assessed the method's viability as an alternative to human visual assessment. A support vector machine based on the image data generated an accuracy rate of 90.2% in discriminating the level of chalkiness, and principal‐components analysis of the image data provided new quantitative variables related to the location and degree of chalkiness with much greater accuracy than was previously possible. These results indicate that image processing may be a useful tool for evaluating the chalkiness of rice in scientific research and breeding programs.