Classification of Thai Rice Seed Cultivars with Image Processing

Abstract

A good quality of rice seed cultivars is the first and most important factor to improve efficiency and reduce cost of rice production. Classification of rice seed from a vast variety of cultivars in Thailand is a challenge task. It requires a sharp visual inspection from only a handful of experts. The evaluation process is tedious and time consuming; moreover, the decision making capabilities of grain inspector can be seriously affected by physical condition such as fatigue, eyesight or even by mental state. The objectives of this research were to study physical features of rice seed cultivars from photographed rice images and to construct an image processing method for classification. Two rice cultivars; RD 49 and RD 6, were used to test the accuracy of classification method. The following percentages of RD 49 rice seeds were mixed into RD 6 seeds at 1, 5, 10, 20 and 50%; successively and the result showed that the rate of successful classification of the two cultivars were 100, 100, 100, 100 and 96%; respectively. It was found that using color filtering and ratios of physical features could help improving the accuracy of classification. The developed method can be conveniently implemented in rice seed cultivars classification machine.