Quality inspection of rice grain using image processing techniques

Abstract

The number of whole kernels, head rice and broken rice are important indicators of the physical quality and value of rice trading in Thailand. Currently, there has been developing image processing technology to determine the number of broken grains to enhance the rice quality inspection process. Therefore, this research aimed to developmentally fabricate a prototype machine by using an image processing method and develop a function for image processing programs to measure the quality of rice grains. The prototype machine was designed with a size of 51x28x30 (width x length x height) cm, with a sorting area (image stage) of 1,400 cm², with a field of view of the camera of 78° at a height of 30 cm. Khao Dawk Mali 105 cultivar at 14.32 %w.b. was chosen as a sample for inspection. The rice grains were separated into 3 groups viz whole kernels (and head rice), broken rice, and small broken C1 by using a standard sorting machine (SATAKE). Then, the three were measured in the three-dimension to create reference image data. The function was created by using the program. Developed functions have a procedure that operates as follows: 1) photo conversion to grayscale, 2) object edge detection, 3) contour writing, 4) border area calculation and axis measurement, 5) object box detection, 6) box counting, and 7) calculation of the characteristics of each group's rice grains as a percentage. The prototype machine was used for rice quality inspection at 5-70 g using the indicator as time and accuracy. The results showed that the prototype machine was run with an average of 8-13 seconds per image. The optimum weight of grains was 5-20 grams per time. It has an average accuracy of 91.63-99.57%. This suggests that the developed prototype machine accurately assessed the quality of rice grains.