DisplayTitle

Rice value in the marketplace is defined by quality grading standards. In addition to defining a classification system, the standards provide an assessment of soundness and quality factors, such as moisture content, infestation, and damage, including heat-damaged kernels, red rice, and chalky kernels. Rice grading according to the standards’ guidelines is often implemented by visual inspection of rice samples, which brings subjective bias to the determination of the corresponding quality grade. Visual inspection of the rice samples is based on comparison with standard rice samples or images, however natural variability is extremely large. Often subtle changes in color, texture, or shape cannot be detected by inspectors and are challenging even for instrumentation using machine vision techniques. QualySense has developed a proprietary high-speed single kernel analyzer, the QSorter Explorer and investigated the possibility of developing an advanced Machine Vision system comprising adaptable illumination combined with an optical subsystem allowing the simultaneous imaging of multiple kernel projections. Color images were acquired, preprocessed and classified with various algorithms. As a result, a method for the reliable detection of chalky kernels has been developed enabling the QSorter Explorer to reach classification accuracies higher than 95% and very low repeatability errors. The method is being developed in collaboration with rice-grading inspection agencies and aims to provide a reliable and fast alternative to current quality inspection and grading methods.