Evaluation and analysis the chalkiness of connected rice kernels based on image processing technology and support vector machine

• We could separate 2–5 connected rice kernels and classify chalky rice kernels from images.
• Chalkiness rate, chalkiness area and chalkiness degree could be detected in this paper.
• We tested 24 indica and japonica rice cultivars with computer vision method.
• The classification accuracy for indica and japonica rice reached 98.5% and 97.6%, respectively.

In order to determine the location and type of rice chalkiness accurately, image processing techniques were adopted to process acquired rice kernel images. Connected rice kernels were separated from each other using a convex point matching method. Chalkiness was extracted according to the differences in grayscale levels between chalky and normal regions in the rice kernel and chalky rice kernels were classified by a support vector machine (SVM). The results showed that 2–5 connected rice kernels could be separated accurately using this method and chalky areas could be extracted. The classification accuracy for indica rice and japonica rice reached 98.5% and 97.6%, respectively, by using SVM. Hence, the measurement results are accurate and reliable, and the presented work provides a theoretical and practical basis for the further application of computer vision technology to chalkiness detection.