Representing the Human Experts Judgment on Quality Indices of White Rice by Image Processing and Artificial Intelligence Techniques

Abstract

In the present study, a grading system based on fuzzy logic was developed to simulate the behavior of an expert in the evaluation and classification of physical properties of rice grains (paddy) for pricing the product. Based on two desired quality indices in this study and the input linguistic variables of fuzzy grading system, 250 samples were prepared with different quality conditions which include all the possible states for the rice grains (paddy). Lighting and imaging were carried out from each 250 samples of rice products in the same condition. Image processing algorithm was conducted to extract geometric features and light intensity of grains and also fuzzy product pricing model was developed in MATLAB software. Fuzzy inference system was designed with the help of fuzzy toolkit. The input variables of the fuzzy system designed in this study were degree of milling (DOM) and percent of broken kernels (PBK) that were obtained as a real numbers of an image processing algorithm. In total, 25 rules of If-Then were formulated with considering the number of inputs’ fuzzy sets. Fuzzy inputs for degree of milling and the percentage of broken kernels were five membership functions of very low, low, medium, high and very high that were selected based on the evaluations conducted from quality of rice production within the rice field factories in north of the country. The results of pricing through fuzzy logic indicated good overall matching with results of product pricing by an expert (overall accuracy of 92 percent).