基于机器视觉的有色稻米品质检测

ppearance quality detection of colored rice using machine vision is the key technology for agricultural production modernization. First, a vision detection system of colored rice infrared images was designed. Red rice was taken as the experimental object, and 20 groups of red rice samples were selected, including 10 groups of red japonica and 10 groups of red indica. A total of 100 colored rice images were collected by visual experiments, of which 50 were red japonica images and 50 were red indica images. Then, linear transformation and threshold segmentation were used to extract the chalkiness regions. The broken rice were screened according to the length of major axis of the ellipse which has the same second-order central moment as the grain. And the counting of the target grains was realized using the connected region labeling algorithm. The experimental results show that the accuracy of detecting the chalky kernel percentage is 95.4%, the accuracy of detecting the chalkiness degree is 97.18%, and the accuracy of detecting broken kernel percentage is 98.4%, indicating that the proposed algorithm has high accuracy, which provides technical support for quality detection and grading of colored rice.